Defect prediction method and system for solder paste printing process

Abstract

The invention discloses a defect prediction method and system for a solder paste printing process, and relates to the technical field of printing defect prediction.The defect prediction method comprises the steps that a time-temperature viscosity attenuation mapping function is constructed, the gradual change mechanism of the solder paste viscosity is explicitly quantized, and real-time coupling with the scraper speed is carried out in an equipment period; the input characteristics are dynamically attached to the actual rheological state of the soldering paste; a hard threshold is applied to model output by means of physical stress constraint, the highest risk can be triggered in time under the high-viscosity and heavy-load limit working conditions, and tin shortage and missing report of a tail-end board card are avoided; compared with a traditional constant viscosity and experience threshold mode, the method has the advantages that frequent line stopping or manual sampling inspection is not needed, on the premise that the production takt is not changed, the defect early warning window at the rear printing section can be circulated in advance for multiple times, and the production line is assisted to complete online pressure compensation and speed limit value adjustment.

Description

Technical Field

[0001] The present invention relates to the technical field of printing defect prediction, and in particular to a defect prediction method and system for a solder paste printing process. Background Art

[0002] In the field of high-end electronic manufacturing, the quality of solder paste printing directly determines the SMT assembly yield. As the size of PCB pads shrinks to the 01005 level, the solder paste printing process is extremely sensitive to fluctuations in process parameters. The current mainstream solution uses an online monitoring system to collect real-time data such as scraper pressure, movement speed, and ambient temperature and humidity, combined with time series models such as LSTM to predict typical defects such as insufficient solder and bridging.

[0003] In long-term continuous printing scenarios, such as automotive electronics production lines with single batches exceeding 5,000 pieces, the characteristic of dynamic viscosity increase caused by continuous evaporation of solder paste solvent has not been effectively modeled. Existing systems generally regard viscosity as a constant parameter, or rely on discrete manual sampling data to correct the model. However, the gradual change of viscosity in the later stage of printing causes a surge in the missed reporting rate of terminal tin deficiency defects, and frequent line stops for inspection disrupt the production rhythm.

[0004] Some solutions incorporate viscosity sensors or infrared spectroscopy, but these sensors require contact with materials that are easily contaminated by solder paste, and non-contact detection is subject to interference from ambient light. Some manufacturers employ dynamic parameter adjustment through reinforcement learning, but due to a lack of physical prior knowledge of viscosity attenuation, convergence efficiency is low. Therefore, a defect prediction solution for the solder paste printing process is urgently needed to address this issue. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a defect prediction method and system for the solder paste printing process to solve the problems of unmodeled viscosity during long-term printing, missed reporting of insufficient solder at the end, and broken beat during line stop detection.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a defect prediction method for a solder paste printing process, which comprises: Step S1, real-time collection of process timing data of the printing press, including scraper pressure, scraper movement speed, ambient temperature and timestamp; Step S2, dynamically generating a viscosity attenuation factor based on the timestamp and ambient temperature using a predefined viscosity attenuation calculation rule; Step S3, performing real-time coupling calculation on the viscosity attenuation factor and the scraper movement speed, and outputting a speed correction parameter; Step S4, inputting the speed correction parameter and the scraper pressure into a pre-trained time series prediction model to output a low tin defect risk level; Step S5: When the risk level exceeds a preset threshold, a real-time warning is triggered and a scraper control instruction is generated.

[0008] In a second aspect, the present invention provides a defect prediction system for a solder paste printing process, comprising: A multi-source sensing module configured to collect scraper pressure, scraper movement speed, and ambient temperature; a viscosity dynamic characterization module, connected to the multi-source sensing module and configured to execute a viscosity decay calculation rule based on a timestamp and an ambient temperature; A real-time coupling analysis unit, inputting and connecting the viscosity dynamic characterization module and the multi-source sensing module, configured to perform coupling operations; A defect prediction engine, having inputs connected to a real-time coupling analysis unit and a multi-source sensing module, configured to run a timing prediction model; A control interface module is configured to parse the risk level and output control instructions to the printing press PLC.

[0009] As a preferred embodiment of the defect prediction method for solder paste printing process described in the present invention, the viscosity attenuation calculation rule in step S2 includes: Establish a time-temperature decay mapping function, in which the temperature weight increases nonlinearly with the continuous printing time; The mapping function output is converted into a dimensionless attenuation coefficient through a linear transformation layer.

[0010] As a preferred embodiment of the defect prediction method for solder paste printing process described in the present invention, in step S2, the viscosity attenuation factor is dynamically generated, and the steps include: When the continuous printing time is When , the time weight tends to saturation exponentially, the duration unit is min, and the growth formula is: , in, is the time weight, dimensionless, is the time weight magnification coefficient, dimensionless, is the time decay rate constant, in units of , is the continuous printing time; The weight is coupled with the real-time ambient temperature to obtain the temperature-time index: , in, is the temperature-time coupling index, in °C, is the ambient temperature in °C, is the reference temperature threshold, in °C; Introducing scale-adaptive linearization, we obtain the dimensionless attenuation coefficient: , in, is the linear scaling factor in units of , The maximum ambient temperature acceptable to the workstation, in °C. is the theoretical maximum time weight, dimensionless; The viscosity attenuation factor is: , in, is the viscosity attenuation factor, is a linear offset term, dimensionless, used to convert Limited to the range of 0-1; when When , hard saturation truncation is performed.

[0011] As a preferred embodiment of the defect prediction method for solder paste printing process described in the present invention, the time series prediction model in step S4 is an LSTM network, and the scraper movement speed data received by its input layer is preprocessed as follows: Multiplying the original speed value by the speed correction parameter; Normalize the product result.

[0012] As a preferred solution of the defect prediction system for solder paste printing process described in the present invention, the real-time coupling analysis unit has a built-in physical constraint submodule, which: Receive viscosity attenuation factor and speed correction parameters; Generate coupling validity flags based on predefined dynamic stress constraints; When the flag is abnormal, the output of the mandatory coverage defect prediction engine is the highest risk level.

[0013] As a preferred embodiment of the defect prediction system for solder paste printing process described in the present invention, in the physical constraint submodule, when the output of the defect prediction engine is the highest risk level, the scraper force is decomposed into normal pressure and shear stress in each sampling cycle, both of which are in Pascals (Pa) as the unified dimension: Define normal pressure as: , in, is the normal pressure, is the current scraper pressure, in N, is the effective width of the scraper, in mm, is the scraper contact length, in mm; Define shear stress as: , in, is the shear stress, in Pa, is the solder paste viscosity-velocity coupling coefficient, in units of , The scraper moving speed after correction of speed correction parameters, in mm. , is the thickness of the solder paste layer, in mm; The dynamic stress index is defined as: , in, is the dynamic stress index, in Pa; Stress ratio and validity sign: , , in, is the stress ratio, dimensionless, is the upper limit of dynamic stress, in Pa, It is the coupling validity mark; The risk coverage logic is expressed as: , in, is the final risk level, is the original risk level output by the LSTM model, The highest risk level defined for the system.

[0014] As a preferred embodiment of the defect prediction method for solder paste printing process described in the present invention, the scraper control instruction in step S5 includes: Pressure compensation value: Increase the scraper pressure proportionally according to the risk level; Speed ​​Limit: Dynamically set the upper limit of movement speed based on the current speed correction parameter.

[0015] As a preferred solution of the defect prediction system for solder paste printing process described in the present invention, the defect prediction engine integrates an online learning module, which: Receive early warning results and actual defect detection feedback; The hidden layer weights of the LSTM network are updated through a sliding window sampling mechanism.

[0016] As a preferred solution of the defect prediction system for solder paste printing process described in the present invention, the multi-source sensing module further includes: Infrared temperature sensor, pointing to the solder paste container area; Timer, synchronously records the start and stop time of printing action; The data fusion unit is configured to align the sampling frequency of the ambient temperature and the time stamp.

[0017] The beneficial effects of the present invention are as follows: by constructing a time-temperature viscosity decay mapping function, the present invention explicitly quantifies the gradual change mechanism of solder paste viscosity and couples it with the scraper speed in real time within the equipment cycle, so that the input characteristics dynamically fit the actual rheological state of the solder paste. Physical stress constraints are then used to impose a hard threshold on the model output, which can immediately trigger the highest risk under high viscosity and heavy load extreme conditions, eliminating the missed reports of insufficient solder on the end board. Compared with traditional constant viscosity and empirical threshold methods, the present invention does not require frequent line stops or manual spot checks, and can advance the defect warning window for the back-end of printing by multiple cycles without changing the production cycle, assisting the production line in completing online pressure compensation and speed limit adjustment. At the same time, the online learning module uses defect feedback to slide and update the LSTM weights, enabling the model to adapt to different batches of solder paste, ambient temperature zones, and equipment aging, significantly improving long-term stability. The overall architecture of the present invention integrates mechanism a priori, data-driven and closed-loop control. On the one hand, it reduces the end-of-batch defect rate and scrap risk, and on the other hand, it reduces unplanned downtime and maintenance time, thereby improving continuous printing yield and equipment utilization, and has significant economic and quality benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 Schematic diagram of the process of defect prediction for solder paste printing process in Example 1.

[0020] Figure 2 Schematic diagram of the framework of the defect prediction system for the solder paste printing process in Example 1. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0024] In the field of high-end electronic manufacturing, the quality of solder paste printing directly determines the SMT assembly yield. As the size of PCB pads shrinks to the 01005 level, the solder paste printing process is extremely sensitive to fluctuations in process parameters. The current mainstream solution uses an online monitoring system to collect real-time data such as scraper pressure, movement speed, and ambient temperature and humidity, combined with time series models such as LSTM to predict typical defects such as insufficient solder and bridging.

[0025] In long-term continuous printing scenarios, such as automotive electronics production lines with single batches exceeding 5,000 pieces, the characteristic of dynamic viscosity increase caused by continuous evaporation of solder paste solvent has not been effectively modeled. Existing systems generally regard viscosity as a constant parameter, or rely on discrete manual sampling data to correct the model. However, the gradual change of viscosity in the later stage of printing causes a surge in the missed reporting rate of terminal tin deficiency defects, and frequent line stops for inspection disrupt the production rhythm.

[0026] Some solutions incorporate viscosity sensors or infrared spectroscopy, but these sensors require contact with materials that are easily contaminated by solder paste, and non-contact detection is subject to interference from ambient light. Some manufacturers employ dynamic parameter adjustment through reinforcement learning, but due to a lack of physical prior knowledge of viscosity attenuation, convergence efficiency is low. Therefore, a defect prediction solution for the solder paste printing process is urgently needed to address this issue.

[0027] Example 1, with reference to Figure 1 and Figure 2 This embodiment provides a defect prediction method for a solder paste printing process, comprising the following steps: Step S1, real-time collection of process timing data of the printing press, including scraper pressure, scraper movement speed, ambient temperature and timestamp; Step S2, dynamically generating a viscosity attenuation factor based on the timestamp and ambient temperature using a predefined viscosity attenuation calculation rule; The viscosity decay calculation rules in step S2 include: Establish a time-temperature decay mapping function, in which the temperature weight increases nonlinearly with the continuous printing time; The mapping function output is converted into a dimensionless attenuation coefficient through a linear transformation layer; In step S2, the viscosity attenuation factor is dynamically generated, and the steps include: When the continuous printing time is When , the time weight tends to saturation exponentially, the duration unit is min, and the growth formula is: , in, is the time weight, dimensionless, is the time weight magnification coefficient, ranging from 0.5 to 2.0, dimensionless, is the time decay rate constant, in units of , is the continuous printing time; The weight is coupled with the real-time ambient temperature to obtain the temperature-time index: , in, is the temperature-time coupling index, in °C, is the ambient temperature in °C, is the reference temperature threshold, in °C; Introducing scale-adaptive linearization, we obtain the dimensionless attenuation coefficient: , in, is the linear scaling factor in units of , The maximum ambient temperature acceptable to the workstation, in °C. is the theoretical maximum time weight, dimensionless; The viscosity attenuation factor is: , in, is the viscosity attenuation factor, is a linear offset term, dimensionless, used to convert Limited to the range of 0-1; when When , hard saturation truncation is performed; Specifically, this step uses the exponential amplification term to highlight the viscosity drop caused by long-term continuous printing, and compresses the temperature-time coupling index to a controllable range through linear scaling, making the attenuation factor dimensionally compatible with the subsequent speed correction, which facilitates rapid network convergence; the parameter and The system can be adjusted based on historical operating conditions, and the linearization coefficient can adapt to different temperature windows. The hard saturation strategy suppresses the numerical overflow caused by extreme operating conditions and maintains the stable operation of the entire prediction chain. Step S3, performing real-time coupling calculation on the viscosity attenuation factor and the scraper movement speed, and outputting a speed correction parameter; Step S4, inputting the speed correction parameter and the scraper pressure into the pre-trained time series prediction model to output the risk level of the low tin defect; The time series prediction model in step S4 is an LSTM network, and the scraper movement speed data received by its input layer is preprocessed as follows: Multiply the original speed value by the speed correction parameter; Normalize the product result; Step S5: When the risk level exceeds a preset threshold, a real-time warning is triggered and a scraper control instruction is generated; The scraper control instructions in step S5 include: Pressure compensation value: Increase the scraper pressure proportionally according to the risk level; Speed ​​Limit: Dynamically set the upper limit of movement speed based on the current speed correction parameter.

[0028] This embodiment further provides a defect prediction system for a solder paste printing process, comprising: A multi-source sensing module configured to collect scraper pressure, scraper movement speed, and ambient temperature; A viscosity dynamic characterization module, connected to the multi-source sensing module, configured to execute viscosity decay calculation rules based on timestamp and ambient temperature; A real-time coupling analysis unit, inputting and connecting the viscosity dynamic characterization module and the multi-source sensing module, configured to perform coupling operations; The real-time coupling analysis unit has a built-in physical constraint submodule, which: Receive viscosity attenuation factor and speed correction parameters; Generate coupling validity flags based on predefined dynamic stress constraints; When the flag is abnormal, the output of the mandatory coverage defect prediction engine is the highest risk level; In the physical constraint submodule, when the output of the defect prediction engine is the highest risk level, the blade force is decomposed into normal pressure and shear stress in each sampling cycle, both of which are in Pascals (Pa): Define normal pressure as: , in, is the normal pressure, is the current scraper pressure, in N, is the effective width of the scraper, in mm, is the scraper contact length, in mm; Define shear stress as: , in, is the shear stress, in Pa, is the solder paste viscosity-velocity coupling coefficient, in units of , The scraper moving speed after correction of speed correction parameters, in mm. , is the thickness of the solder paste layer, in mm; The dynamic stress index is defined as: , in, is the dynamic stress index, in Pa; Stress ratio and validity sign: , , in, is the stress ratio, dimensionless, is the upper limit of dynamic stress, in Pa, It is the coupling validity mark; The risk coverage logic is expressed as: , in, is the final risk level, is the original risk level output by the LSTM model, The highest risk level defined for the system; Specifically, the scraper contact length is introduced and solder paste layer thickness The pressure and speed process quantities are uniformly converted into Pa scale so that the dynamic stress index Intuitively measure the load state of solder paste within the same physical dimension; normal pressure describes the extrusion effect, and shear stress reflects the rheological shear caused by scraper movement. The linear addition of the two is easy to calculate in real time on the PLC; stress ratio is combined with the experimental limit threshold Trigger risk coverage with a hard threshold to avoid misjudgment of the model in extreme load areas; A defect prediction engine, having inputs connected to a real-time coupling analysis unit and a multi-source sensing module, configured to run a timing prediction model; The defect prediction engine integrates an online learning module, which: Receive early warning results and actual defect detection feedback; Update the hidden layer weights of the LSTM network through a sliding window sampling mechanism; The multi-source sensing module also includes: Infrared temperature sensor, pointing to the solder paste container area; Timer, synchronously records the start and stop time of printing action; a data fusion unit configured to align the sampling frequency of the ambient temperature and the timestamp; A control interface module is configured to parse the risk level and output control instructions to the printing press PLC.

[0029] In summary, this embodiment explicitly quantifies the gradual change mechanism of solder paste viscosity by constructing a time-temperature viscosity decay mapping function. This function is then coupled with the scraper speed in real time within the equipment cycle, dynamically adapting the input features to the actual rheological state of the solder paste. Physical stress constraints are then used to impose a hard threshold on the model output, instantly triggering the highest risk under extreme conditions of high viscosity and heavy loads, eliminating missed reports of insufficient solder on the end-of-line boards. Compared with traditional constant viscosity and empirical threshold methods, this invention eliminates the need for frequent line stops or manual spot checks. It can advance the post-printing defect warning window by multiple cycles without changing the production cycle, assisting the production line in completing online pressure compensation and speed limit adjustment. Furthermore, the online learning module uses defect feedback to slidingly update the LSTM weights, enabling the model to adapt to different solder paste batches, ambient temperature zones, and equipment aging, significantly improving long-term stability. The overall architecture integrates mechanism priors, data-driven control, and closed-loop control, reducing end-of-batch defect rates and scrap risks while also reducing unplanned downtime and maintenance time, thereby improving continuous printing yield and equipment utilization, achieving significant economic and quality benefits.

[0030] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A defect prediction method for solder paste printing process, characterized in that: include, Step S1, real-time collection of process timing data of the printing press, including scraper pressure, scraper movement speed, ambient temperature and timestamp; Step S2, dynamically generating a viscosity attenuation factor based on the timestamp and ambient temperature using a predefined viscosity attenuation calculation rule; Step S3, performing real-time coupling calculation on the viscosity attenuation factor and the scraper movement speed, and outputting a speed correction parameter; Step S4, inputting the speed correction parameter and the scraper pressure into a pre-trained time series prediction model to output a low tin defect risk level; Step S5: When the risk level exceeds a preset threshold, a real-time warning is triggered and a scraper control instruction is generated.

2. A defect prediction system for a solder paste printing process, based on the defect prediction method for a solder paste printing process according to claim 1, characterized in that: include: A multi-source sensing module configured to collect scraper pressure, scraper movement speed, and ambient temperature; a viscosity dynamic characterization module, connected to the multi-source sensing module and configured to execute a viscosity decay calculation rule based on a timestamp and an ambient temperature; A real-time coupling analysis unit, inputting and connecting the viscosity dynamic characterization module and the multi-source sensing module, configured to perform coupling operations; A defect prediction engine, having inputs connected to a real-time coupling analysis unit and a multi-source sensing module, configured to run a timing prediction model; A control interface module is configured to parse the risk level and output control instructions to the printing press PLC.

3. The defect prediction method for solder paste printing process according to claim 1, characterized in that: The viscosity decay calculation rules in step S2 include: Establish a time-temperature decay mapping function, in which the temperature weight increases nonlinearly with the continuous printing time; The mapping function output is converted into a dimensionless attenuation coefficient through a linear transformation layer.

4. A defect prediction method for solder paste printing process according to claim 3, characterized in that: In step S2, the viscosity attenuation factor is dynamically generated, and the steps include: When the continuous printing time is When , the time weight tends to saturation exponentially, the duration unit is min, and the growth formula is: ， in, is the time weight, dimensionless, is the time weight magnification coefficient, dimensionless, is the time decay rate constant, in units of , is the continuous printing time; The weight is coupled with the real-time ambient temperature to obtain the temperature-time index: ， in, is the temperature-time coupling index, in °C, is the ambient temperature in °C, is the reference temperature threshold, in °C; Introducing scale-adaptive linearization, we obtain the dimensionless attenuation coefficient: ， in, is the linear scaling factor in units of , The maximum ambient temperature acceptable to the workstation, in °C. is the theoretical maximum time weight, dimensionless; The viscosity attenuation factor is: ， in, is the viscosity attenuation factor, is a linear offset term, dimensionless, used to convert Limited to the range of 0-1; when When , hard saturation truncation is performed.

5. The defect prediction method for solder paste printing process according to claim 1, characterized in that: The time series prediction model in step S4 is an LSTM network, and the scraper movement speed data received by its input layer is preprocessed as follows: Multiplying the original speed value by the speed correction parameter; Normalize the product result.

6. A defect prediction system for solder paste printing process according to claim 2, characterized in that: The real-time coupling analysis unit has a built-in physical constraint submodule, which: Receive viscosity attenuation factor and speed correction parameters; Generate coupling validity flags based on predefined dynamic stress constraints; When the flag is abnormal, the output of the mandatory coverage defect prediction engine is the highest risk level.

7. A defect prediction system for solder paste printing process according to claim 6, characterized in that: In the physical constraint submodule, when the output of the defect prediction engine is the highest risk level, the blade force is decomposed into normal pressure and shear stress in each sampling period, both of which are in Pascals (Pa): Define normal pressure as: ， in, is the normal pressure, is the current scraper pressure, in N, is the effective width of the scraper, in mm, is the scraper contact length, in mm; Define shear stress as: ， in, is the shear stress, in Pa, is the solder paste viscosity-velocity coupling coefficient, in units of , The scraper moving speed after correction of speed correction parameters, in mm. , is the thickness of the solder paste layer, in mm; The dynamic stress index is defined as: ， in, is the dynamic stress index, in Pa; Stress ratio and validity sign: ， ， in, is the stress ratio, dimensionless, is the upper limit of dynamic stress, in Pa, It is the coupling validity mark; The risk coverage logic is expressed as: ， in, is the final risk level, is the original risk level output by the LSTM model, The highest risk level defined for the system.

8. The defect prediction method for solder paste printing process according to claim 1, characterized in that: The scraper control instructions in step S5 include: Pressure compensation value: Increase the scraper pressure proportionally according to the risk level; Speed ​​Limit: Dynamically set the upper limit of movement speed based on the current speed correction parameter.

9. A defect prediction system for solder paste printing process according to claim 2, characterized in that: The defect prediction engine integrates an online learning module, which: Receive early warning results and actual defect detection feedback; The hidden layer weights of the LSTM network are updated through a sliding window sampling mechanism.

10. A defect prediction system for solder paste printing process according to claim 2, characterized in that: The multi-source sensing module further includes: Infrared temperature sensor, pointing to the solder paste container area; Timer, synchronously records the start and stop time of printing action; The data fusion unit is configured to align the sampling frequency of the ambient temperature and the time stamp.

Images