Computer mainboard welding temperature correction control method based on machine learning

Abstract

The invention discloses a computer mainboard welding temperature correction control method based on machine learning, which relates to the technical field of intelligent control, and comprises the following steps: predicting a reference welding temperature curve by using a machine learning model based on historical operation data of reflow welding of a computer mainboard; dividing a thermal management area based on the thermal characteristics of the component; determining an optimal welding temperature curve; on the basis of the divided heat management areas, regionalization adjustment is conducted on the optimal welding temperature curve, and temperature area parameter fine adjustment strategies for the different heat management areas are generated; and based on the optimal welding temperature curve and the fine adjustment strategy, a final temperature control instruction set is synthesized, and the reflow soldering furnace is controlled for batch production. The method has the advantages that through fusion of machine learning and real-time data driving optimization, on the premise that the welding reliability is guaranteed, the production efficiency and energy consumption are intelligently balanced, and the adaptability and the intelligent level of a production line are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, in particular to a computer mainboard welding temperature correction control method based on machine learning. BACKGROUND

[0002] In the field of computer mainboard reflow soldering process, the traditional method mainly relies on fixed temperature curve or limited experience adjustment to control the reflow soldering furnace. These methods usually adopt "one-size-fits-all" temperature setting, which is difficult to adapt to the diversity of component layout, packaging type and heat capacity on different models of mainboards. Due to the lack of detailed consideration of complex heat interaction in the welding process, it often leads to unstable welding quality, resulting in defects such as cold soldering, cold soldering or thermal damage, which not only affects product yield, but also restricts production efficiency and cost control. With the development of electronic products towards high density and miniaturization, the thermal characteristics of components on the mainboard become more and more different. This extensive temperature control method has been difficult to meet the requirements of high reliability manufacturing.

[0003] To overcome the above limitations, the industry has begun to explore the introduction of data-driven methods. However, the existing preliminary intelligent solutions mostly focus on simple statistical analysis of historical data or single target optimization, and fail to deeply integrate the specific physical layout of the mainboard, the real-time thermal management needs of the components and the welding process parameters for adaptive adjustment. When dealing with new materials and new board types, it lacks effective and rapid response and self-learning ability, and the determination of process window is still rigid, making it difficult to achieve dynamic optimal balance between quality, efficiency and energy consumption. Therefore, there is an urgent need for an intelligent control method that can comprehensively consider multiple factors and continuously optimize itself to achieve precise, efficient and adaptive setting of the welding temperature curve. SUMMARY

[0004] To solve the above technical problems, a computer mainboard welding temperature correction control based on machine learning is provided, which solves at least one of the technical problems mentioned in the background technology.

[0005] To achieve the above purposes, the technical solution adopted by the present application is:

[0006] A computer mainboard welding temperature correction control method based on machine learning, comprising:

[0007] Based on the historical operation data of computer mainboard reflow soldering, a machine learning model is used to predict a reference welding temperature curve, which includes the set temperature of multiple temperature zones of the reflow soldering furnace;

[0008] Collecting image information of the computer mainboard to be welded, identifying the layout of components on the mainboard, and dividing the thermal management area based on the thermal characteristics of the components;

[0009] Based on the aforementioned reference welding temperature curve, small-batch trial production was carried out, and post-weld quality data was collected.

[0010] Based on the quality data, the parameters of the reference welding temperature curve are iteratively optimized to determine an optimal welding temperature curve.

[0011] Based on the defined heat management regions, the optimal welding temperature curve is adjusted regionally to generate a fine-tuning strategy for temperature zone parameters for different heat management regions.

[0012] Based on the optimal welding temperature curve and the fine-tuning strategy, a final temperature control instruction set is synthesized to control the reflow oven for mass production.

[0013] Preferably, the step of using a machine learning model to predict a baseline soldering temperature profile based on historical reflow soldering data of computer motherboards specifically includes:

[0014] Obtain temperature curves from multiple samples in historical successful cases;

[0015] Key parameters for each temperature zone are extracted from each sample temperature curve to form a sample parameter set;

[0016] The sample parameter set was cleaned using the Grubbs criterion to remove outlier data;

[0017] The standard sample parameter set after cleaning is weighted and averaged. The weights are determined based on the similarity of the motherboard design. The calculation result is the initial parameter of the reference welding temperature curve.

[0018] Preferably, the step of iteratively optimizing the parameters of the reference welding temperature curve based on the quality data to determine an optimal welding temperature curve specifically includes:

[0019] Select several representative test soldering points on the computer motherboard to be soldered;

[0020] After trial welding and trial production of the test welding points using the reference welding temperature, welding quality data is obtained through automatic optical inspection (AOI) and X-ray inspection. The data includes weld point brightness, shape, cold welds, and component offset information.

[0021] If the welding quality of the trial welding point is not up to standard, the reference welding temperature plus the temperature difference value shall be used as the updated reference welding temperature. After trial welding and trial production at the updated reference welding temperature, the welding quality shall be re-inspected. This process shall be repeated until the lowest temperature value that makes the welding quality up to standard is found, and this temperature value shall be recorded as the lower limit.

[0022] Once the lower temperature limit is found, continue to increase the temperature difference for trial welding and trial production until the welding quality begins to decline or thermal damage occurs. Record the temperature value before stopping as the upper temperature limit.

[0023] The optimal welding temperature is calculated based on the lower and upper temperature limits, using an optimization algorithm.

[0024] Preferably, the calculation of the optimal welding temperature based on the lower and upper temperature limits using an optimization algorithm specifically involves:

[0025] The obtained effective welding temperature range is recorded as the process window, where the lower temperature limit is... The lower boundary of the window, and the upper temperature limit. The upper boundary of the window;

[0026] Based on the thermal sensitivity distribution of components on the motherboard and the soldering reliability requirements, using the formula... Calculate the temperature buffer value, where k is the reliability coefficient determined based on historical process data. The value range is from 0.1 to 0.2;

[0027] The optimal welding temperature is calculated using an optimization algorithm. The calculation formula is as follows:

[0028]

[0029] in, To optimize the weighting coefficients, a machine learning model was used to perform multi-objective optimization on welding quality, energy consumption, and efficiency to determine the weighting coefficients.

[0030] Preferably, the method for determining the optimization weight coefficients is as follows:

[0031] For each optimization weight coefficient, the calculated Determine the optimal conveyor belt speed, which is the speed at which the conveyor belt speed is determined. The fastest conveyor belt speed that satisfies welding quality requirements;

[0032] Calculated based on each optimization weight coefficient Trial production was conducted using the corresponding optimal conveyor belt speed;

[0033] The number of computer motherboards soldered per unit time is used as an efficiency indicator.

[0034] The energy consumption index is calculated by dividing the energy consumption per unit time by the number of computer motherboards soldered per unit time.

[0035] Welding quality is evaluated by weighted comprehensive assessment of solder joint yield and component offset, which serves as the quality indicator.

[0036] Based on efficiency, energy consumption, and quality indicators, the TOPSIS method is used to select the optimal [result]. and the optimal The corresponding optimization weight coefficients are used as the final determined optimization weight coefficients.

[0037] Preferably, the step of acquiring image information of the computer motherboard to be soldered, identifying the component layout on the motherboard, and dividing the thermal management area based on the thermal characteristics of the components specifically includes:

[0038] The motherboard topology image is acquired using a high-resolution camera, and then noise reduction and contrast enhancement are performed.

[0039] Deep learning object detection algorithms are used to identify various components in images and record their type, package size, and precise coordinates on the motherboard.

[0040] Based on the known thermal characteristics of the components, the motherboard surface is divided into a high heat capacity area, a heat-sensitive area, and a standard soldering area, and a corresponding thermal management label is generated for each area.

[0041] Preferably, the step of regionally adjusting the optimal welding temperature curve based on the divided heat management regions to generate a fine-tuning strategy for temperature zone parameters in different heat management regions specifically includes:

[0042] Based on the geometric model of the motherboard and the division of the thermal management area, thermal field simulation is performed to predict the actual temperature of different thermal management areas during the reflow process.

[0043] Compare the simulation results with the target temperature curve, calculate the required increase in temperature compensation for high heat capacity regions, and calculate the required decrease in temperature and / or slowing down the heating rate for heat-sensitive regions.

[0044] The calculated compensation values ​​for each region are integrated into the target temperature curve to generate a final temperature control curve with spatially differentiated parameters.

[0045] Preferably, the step of synthesizing the final temperature control instruction set based on the optimal welding temperature curve and the fine-tuning strategy to control the reflow oven for mass production specifically includes:

[0046] During mass production, thermocouples and vision systems are used to collect real-time images of the actual temperature curve inside the furnace and images of key welding points.

[0047] The real-time data is compared with the expected final temperature curve. If the deviation continues to exceed the tolerance range, an adjustment decision is triggered.

[0048] Based on the decision results, the corresponding temperature zone setpoint of the reflow oven is automatically and gradually corrected to restore the process to a controlled state.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0050] This invention achieves accurate prediction and adaptive adjustment of reflow soldering temperature profiles for computer motherboards by integrating machine learning and real-time data-driven optimization. It significantly improves the consistency of soldering quality and yield rate. By intelligently dividing thermal management zones and implementing differentiated temperature control strategies, it effectively avoids defects such as cold solder joints, poor soldering, and component thermal damage caused by uneven heat distribution. Simultaneously, this method, through a multi-objective optimization algorithm, intelligently balances production efficiency and energy consumption while ensuring soldering reliability, achieving a comprehensive optimization of quality, efficiency, and cost, and significantly enhancing the adaptability and intelligence level of the production line. Attached Figure Description

[0051] Figure 1 This is a flowchart of the machine learning-based computer motherboard soldering temperature correction and control method proposed in this solution.

[0052] Figure 2 This is a flowchart of the method proposed in this scheme for predicting a baseline welding temperature profile using a machine learning model.

[0053] Figure 3 Flowchart of the method for determining an optimal welding temperature profile through iterative optimization proposed in this scheme;

[0054] Figure 4 This is a flowchart of the method for dividing the thermal management area based on the thermal characteristics of components proposed in this solution;

[0055] Figure 5 The flowchart of the method proposed in this solution for generating temperature zone parameter fine-tuning strategies for different thermal management zones is shown below;

[0056] Figure 6 This is a flowchart illustrating the method for controlling reflow ovens for mass production as proposed in this scheme.

[0057] Figure 7 This is an architecture diagram of the electronic devices in this solution;

[0058] Figure 8 This is a schematic diagram of the computer-readable storage medium structure in this scheme.

[0059] The numbers on the map are:

[0060] 500 - Electronic device; 501 - Bus; 502 - CPU; 503 - ROM; 504 - RAM; 505 - Communication port; 506 - Input / output component; 507 - Hard disk; 508 - User interface; 600 - Computer-readable storage medium. Detailed Implementation

[0061] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0062] Reference Figure 1 As shown, a machine learning-based method for correcting and controlling the soldering temperature of a computer motherboard includes:

[0063] Based on historical reflow soldering data of computer motherboards, a machine learning model is used to predict a baseline soldering temperature profile. This profile includes the set temperatures for multiple zones of the reflow oven. By mining a vast amount of historical successful cases through the machine learning model, the system can automatically learn the complex mapping relationship between different types of motherboards and their optimal soldering temperature profiles. This allows for the rapid generation of a scientific and reasonable baseline temperature profile as an optimization starting point when facing new motherboard types. This replaces the traditional manual setting method relying on engineers' personal experience, significantly improving the accuracy and efficiency of initial settings and laying a reliable data foundation for subsequent fine-tuning.

[0064] The process involves acquiring image information of the computer motherboard to be soldered, identifying the component layout on the motherboard, and dividing the motherboard into thermal management zones based on the thermal characteristics of the components. Computer vision technology is used to automatically identify the components on the motherboard and, based on their physical characteristics such as heat capacity and heat sensitivity (e.g., large BGA packages are high heat capacity areas, while crystal oscillators and connectors are heat-sensitive areas), the motherboard is divided into different thermal management zones. This marks a shift in concept from "uniform temperature control for the entire board" to "precise temperature control by zone," providing a crucial basis for subsequent differentiated temperature adjustments for the specific thermal needs of different areas, and preventing quality defects caused by uneven heat distribution from the outset.

[0065] Based on the aforementioned benchmark welding temperature curve, small-batch trial production was conducted, and post-weld quality data was collected. Through this trial production, the benchmark curve was tested under actual working conditions. Using non-destructive testing methods such as automated optical inspection (AOI) and X-ray, key quality data such as weld joint morphology and weld strength were systematically collected, transforming abstract process parameters into concrete and quantifiable quality information. This process provides real and objective data feedback for subsequent optimization iterations, ensuring that the optimization direction is always closely related to the actual welding effect.

[0066] Based on the quality data, the parameters of the baseline welding temperature curve are iteratively optimized to determine an optimal welding temperature curve. The machine learning model deeply analyzes the causal relationship between the quality data and temperature parameters, automatically diagnoses the root causes of defects, and proposes specific parameter adjustment suggestions accordingly. Through multiple iterations of optimization, the system can dynamically converge to an "optimal welding temperature curve" that performs best under given conditions, thereby achieving autonomous and intelligent improvement of process parameters, rather than relying on manual trial and error.

[0067] Based on the defined thermal management regions, the optimal welding temperature curve is regionally adjusted to generate fine-tuning strategies for temperature zone parameters in different thermal management regions. Building upon the global optimal curve and combining the previously defined thermal management regions, thermal simulation calculations or rule-based reasoning are used to dynamically compensate heat for high-heat-capacity regions to ensure sufficient welding, and to appropriately reduce the temperature or slow down the heating rate for heat-sensitive regions to provide protection. This generates a spatially differentiated fine-tuning strategy, enabling a single furnace temperature curve to intelligently adapt to the differentiated thermal requirements of different locations on the motherboard, achieving a unification of global and local optimization.

[0068] Based on the optimal welding temperature curve and the fine-tuning strategy, a final temperature control command set is synthesized to control the reflow oven for mass production. The global optimal curve and the regional fine-tuning strategy are integrated and compiled into a control command set that the reflow oven can directly execute, containing parameters such as the set temperature for each temperature zone and the conveyor belt speed. During mass production, closed-loop monitoring and fine-tuning can be performed using real-time temperature sensor data to ensure the stability and reliability of the production process. This ultimately completes a full automated closed loop from data perception and intelligent decision-making to physical execution, achieving a high degree of intelligence and automation in welding production.

[0069] Reference Figure 2 As shown, the prediction of a baseline soldering temperature profile based on historical reflow soldering data of computer motherboards using a machine learning model specifically includes:

[0070] Obtain temperature curves from multiple samples in historical successful cases;

[0071] Key parameters for each temperature zone are extracted from each sample temperature curve to form a sample parameter set;

[0072] The sample parameter set was cleaned using the Grubbs criterion to remove outlier data;

[0073] The formula for Grubbs' criterion is as follows:

[0074]

[0075] In the formula, For the i-th sample parameter, The average of all sample parameters. The standard deviation of all sample parameters, The Grubbs number is obtained by looking up the Grubbs table.

[0076] The standard sample parameter set after cleaning is weighted and averaged. The weights are determined based on the similarity of the motherboard design. The calculation result is the initial parameter of the reference welding temperature curve.

[0077] This process transforms historical experience data into a reliable and scientific initial process baseline. Through systematic data preprocessing, it ensures the accuracy of baseline predictions: First, multi-dimensional parameters extracted from successful cases form the learning foundation; second, statistical methods such as the Grubbs criterion are applied to rigorously eliminate outliers, effectively avoiding interference from flawed historical data and guaranteeing data quality; the most crucial step is the introduction of "motherboard design similarity" as a weighted average, which makes the generation of the baseline curve no longer a simple numerical average, but rather possesses intelligent correlation—the more similar a historical motherboard is to the current motherboard to be produced in terms of size, number of layers, and main chip types, the greater the weight of its process parameters on the initial baseline. This is equivalent to digitally inheriting the valuable experience of senior engineers in "referencing similar boards to set the furnace temperature for the new board," but the process is faster, more objective, and more quantifiable.

[0078] The generated baseline temperature profile serves as a high-quality starting point for optimization. It significantly narrows the search range for subsequent trial production optimization, avoiding the substantial trial-and-error costs and time consumption associated with starting from scratch or relying entirely on manual experience. This accelerates the introduction of new products and improves the stability and convergence efficiency of the entire optimization process. It effectively reduces the probability of first-piece rejection or potential quality risks due to improper initial parameter settings, laying a solid data foundation for producing highly reliable products.

[0079] Reference Figure 3 As shown, the step of iteratively optimizing the parameters of the baseline welding temperature curve based on the quality data to determine an optimal welding temperature curve specifically includes:

[0080] Select several representative test soldering points on the computer motherboard to be soldered;

[0081] After trial welding and trial production of the test welding points using the reference welding temperature, welding quality data is obtained through automatic optical inspection (AOI) and X-ray inspection. The data includes weld point brightness, shape, cold welds, and component offset information.

[0082] If the welding quality of the trial welding point is not up to standard, the reference welding temperature plus the temperature difference value shall be used as the updated reference welding temperature. After trial welding and trial production at the updated reference welding temperature, the welding quality shall be re-inspected. This process shall be repeated until the lowest temperature value that makes the welding quality up to standard is found, and this temperature value shall be recorded as the lower limit.

[0083] Once the lower temperature limit is found, continue to increase the temperature difference for trial welding and trial production until the welding quality begins to decline or thermal damage occurs. Record the temperature value before stopping as the upper temperature limit.

[0084] The optimal welding temperature is calculated based on the lower and upper temperature limits, using an optimization algorithm.

[0085] The calculation of the optimal welding temperature based on the lower and upper temperature limits using an optimization algorithm is as follows:

[0086] The obtained effective welding temperature range is recorded as the process window, where the lower temperature limit is... The lower boundary of the window, and the upper temperature limit. The upper boundary of the window;

[0087] Based on the thermal sensitivity distribution of components on the motherboard and the soldering reliability requirements, using the formula... Calculate the temperature buffer value, where k is the reliability coefficient determined based on historical process data. The value range is from 0.1 to 0.2;

[0088] The optimal welding temperature is calculated using an optimization algorithm. The calculation formula is as follows:

[0089]

[0090] in, To optimize the weighting coefficients, a machine learning model was used to perform multi-objective optimization on welding quality, energy consumption, and efficiency to determine the weighting coefficients.

[0091] The design concept for the optimal welding temperature is as follows:

[0092] and This is the critical point where the welding quality is "just acceptable." Working at this temperature is extremely risky; even the slightest fluctuation can lead to defects. Therefore, a temperature buffer is introduced to move the setpoint a safe distance from the dangerous edge, thereby establishing a safe operating range. In this way, even if there are normal fluctuations in the production process, the temperature can be maintained within the specified range, ensuring the reliability of the welding quality;

[0093] α is a coefficient dynamically calculated by a machine learning model, typically ranging from [0, 1]. It's the "intelligence" of the entire formula, used for precise positioning within the safe optimization range. A smaller α means selecting the lower limit of safe operation. This is suitable for scenarios that are highly sensitive to heat and require the lowest possible temperature, emphasizing component protection and minimal energy consumption. A larger α means selecting the upper limit of safe operation. This is suitable for scenarios with high welding difficulty and requiring higher heat capacity to ensure quality, emphasizing welding strength and efficiency.

[0094] Specifically, the method for determining the optimization weight coefficients is as follows:

[0095] For each optimization weight coefficient, the calculated Determine the optimal conveyor belt speed, which is the speed at which the conveyor belt speed is determined. The fastest conveyor belt speed that satisfies welding quality requirements;

[0096] Calculated based on each optimization weight coefficient Trial production was conducted using the corresponding optimal conveyor belt speed;

[0097] The number of computer motherboards soldered per unit time is used as an efficiency indicator.

[0098] The energy consumption index is calculated by dividing the energy consumption per unit time by the number of computer motherboards soldered per unit time.

[0099] Welding quality is evaluated by weighted comprehensive assessment of solder joint yield and component offset, which serves as the quality indicator.

[0100] Based on efficiency, energy consumption, and quality indicators, the TOPSIS method is used to select the optimal [result]. and the optimal The corresponding optimization weight coefficients are used as the final determined optimization weight coefficients.

[0101] A systematic optimization framework is introduced. Specifically, it doesn't seek the optimal temperature or fastest speed in isolation, but recognizes that temperature (determined by α) and conveyor belt speed are coupled variables jointly determining "heat input" and "production cycle time." Through the linked rule of "finding the fastest speed that meets quality requirements for each α," this method automatically explores the Pareto front in the temperature-speed parameter space—the set of solutions that cannot improve one objective without compromising another. Subsequently, using the TOPSIS method, a multi-criteria decision-making tool, it intelligently selects the optimal solution from the Pareto front that best aligns with the company's overall interests, based on different emphases on quality, efficiency, and energy consumption in actual production. This process scientifically quantifies the traditionally fuzzy concept of "process optimization," providing a general and efficient paradigm for solving complex industrial optimization problems involving multiple coupled objectives.

[0102] By constructing a multi-objective optimization system integrating quality, efficiency, and energy consumption, and combining it with the TOPSIS decision-making method, global intelligent optimization of welding process parameters has been achieved. The core benefit of this method lies in its complete transformation of the traditional single-point debugging or experience-based parameter setting mode. By linking and coordinating the optimization of conveyor belt speed and welding temperature, it ensures that the optimal process window with the highest production efficiency and lowest energy consumption can be automatically found while meeting welding quality requirements. This not only significantly improves the consistency of welding quality and production efficiency, and greatly reduces energy consumption and production costs, but more importantly, through a scientifically quantified decision-making mechanism, it transforms process parameter setting from an "experience-based skill" into a reproducible and traceable standardized process, greatly enhancing the stability, reliability, and intelligence level of mass production.

[0103] Reference Figure 4 As shown, the process of acquiring image information of the computer motherboard to be soldered, identifying the component layout on the motherboard, and dividing the thermal management area based on the thermal characteristics of the components specifically includes:

[0104] The motherboard topology image is acquired using a high-resolution camera, and then noise reduction and contrast enhancement are performed.

[0105] Deep learning object detection algorithms are used to identify various components in images and record their type, package size, and precise coordinates on the motherboard.

[0106] Based on the known thermal characteristics of the components, the motherboard surface is divided into a high heat capacity area, a heat-sensitive area, and a standard soldering area, and a corresponding thermal management label is generated for each area.

[0107] By integrating visual recognition technology with materials thermodynamics, an intelligent self-identification method for thermal management areas has been achieved. The "known thermal characteristic parameters of components" form an updatable knowledge base, encompassing not only intuitive factors like component package size but also deeper physical properties such as specific heat capacity, thermal conductivity, and maximum temperature resistance. By matching deep learning's target detection capabilities with the thermal characteristic parameter database, the system transcends traditional image recognition, achieving accurate prediction and regional division of the thermal load distribution on the PCB board. This division is not a simple geometric partition but a functional partition based on thermal characteristics, providing a crucial prerequisite for subsequent differentiated and precise temperature control. Through automated image recognition and thermal region division, precise and proactive thermal management in the reflow soldering process is achieved. This method can accurately identify high-heat-capacity areas (such as large BGA chips) and heat-sensitive areas (such as micro-crystal oscillators) on the motherboard, enabling targeted design during temperature profile development. Its beneficial effects are mainly reflected in three aspects: First, it effectively avoids local cold welding or overheating damage caused by differences in heat capacity, significantly improving the welding yield; second, it transforms the qualitative judgment that traditionally relies on engineers' experience into quantitative zoning based on data-driven approaches, reducing reliance on manual skills and improving the consistency and repeatability of the process; third, it lays a precise data foundation for the subsequent implementation of differentiated temperature zone control strategies, making precise heating with "one policy for one location" possible, and optimizing welding quality and energy efficiency as a whole.

[0108] Reference Figure 5 As shown, the process of adjusting the optimal welding temperature curve based on the defined thermal management regions to generate a fine-tuning strategy for temperature zone parameters in different thermal management regions specifically includes:

[0109] Based on the geometric model of the motherboard and the division of the thermal management area, thermal field simulation is performed to predict the actual temperature of different thermal management areas during the reflow process.

[0110] Compare the simulation results with the target temperature curve, calculate the required increase in temperature compensation for high heat capacity regions, and calculate the required decrease in temperature and / or slowing down the heating rate for heat-sensitive regions.

[0111] The calculated compensation values ​​for each region are integrated into the target temperature curve to generate a final temperature control curve with spatially differentiated parameters.

[0112] A closed-loop technical path was established, from the division of thermal management zones to specific temperature control strategies. The thermal field simulation stage uses computer numerical simulation to pre-enact the heat conduction behavior of the entire reflow soldering process in a virtual space, essentially performing a "digital twin" test before actual production. By comparing the simulation results with the target curve, the system can quantify the precise compensation required for each thermal management zone. This process essentially transforms engineers' practical experience into a calculable and reusable algorithm model. The resulting differentiated temperature control curves are essentially a set of intelligent control instructions that consider the two-dimensional thermal characteristics of the motherboard, enabling a single reflow oven to achieve customized heating for different areas. This embodies the core concept of "precise control" in intelligent manufacturing. Through thermal field simulation and regional compensation technology, precise spatial control of the temperature field during reflow soldering is achieved, effectively solving the soldering defects caused by local thermal capacity differences on the motherboard. This method elevates traditional "one-size-fits-all" temperature control to "on-demand" precise thermal management, proactively predicting and compensating for uneven heat distribution. Its core benefits lie in significantly improving the consistency of welding quality, avoiding the risks of cold welding in high heat capacity areas and overheating damage in heat-sensitive areas, and reducing energy waste caused by overheating by optimizing heat distribution, thus achieving a dual improvement in quality and energy efficiency.

[0113] Reference Figure 6 As shown, the process of synthesizing the final temperature control command set based on the optimal welding temperature curve and the fine-tuning strategy to control the reflow oven for mass production specifically includes:

[0114] During mass production, thermocouples and vision systems are used to collect real-time images of the actual temperature curve inside the furnace and images of key welding points.

[0115] The real-time data is compared with the expected final temperature curve. If the deviation continues to exceed the tolerance range, an adjustment decision is triggered.

[0116] Based on the decision results, the corresponding temperature zone setpoint of the reflow oven is automatically and gradually corrected to restore the process to a controlled state.

[0117] Thermocouples (sensing physical temperature signals) and a vision system (sensing welding quality) act as "sensors," continuously monitoring the production process. The process of comparing real-time data with expected values ​​(i.e., the "optimal welding temperature profile") is a "brain decision-making" process based on preset rules (such as persistent deviations exceeding limits) that ensure the system intervenes only when necessary, avoiding over-adjustment. Finally, "small-amplitude, gradual" corrections to the temperature setpoint are a classic method in control theory, aiming to smoothly and robustly bring the production process back to a controlled state, preventing system oscillations caused by overly aggressive adjustments. This closed-loop design gives the static "optimal process parameters" dynamic adaptability, thus endowing the production line with higher intelligence and robustness. By introducing real-time data acquisition and a closed-loop feedback control mechanism, the reflow soldering process has achieved a leap from static parameter control to dynamic adaptive optimization. This technical solution can effectively cope with process drift caused by environmental fluctuations, material batch differences, or equipment aging during production, ensuring that each motherboard is soldered within the optimal process window. Its core value lies in significantly improving the consistency and stability of product quality in large-scale mass production, minimizing batch defects caused by process deviations, and reducing reliance on external human intervention through automated process regulation. It is a key technological guarantee for achieving the goals of intelligent manufacturing and "zero-defect" production.

[0118] Furthermore, the method according to the embodiments of this application can also be achieved by means of... Figure 7 The architecture of the electronic device shown is used to implement this. For example... Figure 7 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store a machine learning-based computer motherboard soldering temperature correction control method provided in this application. The electronic device 500 may also include a user interface 508. Of course, Figure 7 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 7 One or more components in the illustrated electronic device.

[0119] Figure 8 This is a schematic diagram of a computer-readable storage medium structure provided in one embodiment of this application. Figure 8The diagram illustrates a computer-readable storage medium 600 according to one embodiment of this application. The computer-readable storage medium 600 stores computer-readable instructions. When executed by a processor, the computer-readable instructions can perform a machine learning-based computer motherboard soldering temperature correction control method according to an embodiment of this application, as described with reference to the above figures. The computer-readable storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0120] In summary, the advantages of this invention are as follows: By integrating machine learning and real-time data-driven optimization, it achieves accurate prediction and adaptive adjustment of the reflow soldering temperature profile for computer motherboards. This significantly improves the consistency of soldering quality and yield. By intelligently dividing the thermal management zone and implementing differentiated temperature control strategies, it effectively avoids defects such as cold solder joints, poor soldering, and component thermal damage caused by uneven heat distribution. Furthermore, through a multi-objective optimization algorithm, this method intelligently balances production efficiency and energy consumption while ensuring soldering reliability, achieving a comprehensive optimization of quality, efficiency, and cost, and significantly enhancing the adaptability and intelligence level of the production line.

[0121] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning, characterized in that, include: Based on historical operation data of reflow soldering of computer motherboards, a machine learning model is used to predict a baseline soldering temperature profile, which includes the set temperatures of multiple temperature zones in the reflow oven. The image information of the computer motherboard to be soldered is acquired, the layout of the components on the motherboard is identified, and the thermal management area is divided based on the thermal characteristics of the components. Based on the aforementioned reference welding temperature curve, small-batch trial production was carried out, and post-weld quality data was collected. Based on the quality data, the parameters of the reference welding temperature curve are iteratively optimized to determine an optimal welding temperature curve. Based on the defined heat management regions, the optimal welding temperature curve is adjusted regionally to generate a fine-tuning strategy for temperature zone parameters for different heat management regions. Based on the optimal welding temperature curve and the fine-tuning strategy, a final temperature control instruction set is synthesized to control the reflow oven for mass production.

2. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 1, characterized in that, The prediction of a baseline soldering temperature profile based on historical reflow soldering data from computer motherboards, using a machine learning model, specifically includes: Obtain temperature curves from multiple samples in historical successful cases; Key parameters for each temperature zone are extracted from each sample temperature curve to form a sample parameter set; The sample parameter set was cleaned using the Grubbs criterion to remove outlier data; The standard sample parameter set after cleaning is weighted and averaged. The weights are determined based on the similarity of the motherboard design. The calculation result is the initial parameter of the reference welding temperature curve.

3. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 1, characterized in that, The step of iteratively optimizing the parameters of the baseline welding temperature curve based on the quality data to determine an optimal welding temperature curve specifically includes: Select several representative test soldering points on the computer motherboard to be soldered; After trial welding and trial production of the test welding points using the reference welding temperature, welding quality data is obtained through automatic optical inspection (AOI) and X-ray inspection. The data includes weld point brightness, shape, cold welds, and component offset information. If the welding quality of the trial welding point is not up to standard, the reference welding temperature plus the temperature difference value shall be used as the updated reference welding temperature. After trial welding and trial production at the updated reference welding temperature, the welding quality shall be re-inspected. This process shall be repeated until the lowest temperature value that makes the welding quality up to standard is found, and this temperature value shall be recorded as the lower limit. Once the lower temperature limit is found, continue to increase the temperature difference for trial welding and trial production until the welding quality begins to decline or thermal damage occurs. Record the temperature value before stopping as the upper temperature limit. The optimal welding temperature is calculated based on the lower and upper temperature limits, using an optimization algorithm.

4. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 3, characterized in that, The calculation of the optimal welding temperature based on the lower and upper temperature limits using an optimization algorithm is as follows: The obtained effective welding temperature range is recorded as the process window, where the lower temperature limit is... The lower boundary of the window, and the upper temperature limit. The upper boundary of the window; Based on the thermal sensitivity distribution of components on the motherboard and the soldering reliability requirements, using the formula... Calculate the temperature buffer value, where k is the reliability coefficient determined based on historical process data. The value range is from 0.1 to 0.2; The optimal welding temperature is calculated using an optimization algorithm. The calculation formula is as follows: in, To optimize the weighting coefficients, a machine learning model was used to perform multi-objective optimization on welding quality, energy consumption, and efficiency to determine the weighting coefficients.

5. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 4, characterized in that, The method for determining the optimization weight coefficients is as follows: For each optimization weight coefficient, the calculated Determine the optimal conveyor belt speed, which is the speed at which the conveyor belt speed is determined. The fastest conveyor belt speed that satisfies welding quality requirements; Calculated based on each optimization weight coefficient Trial production was conducted using the corresponding optimal conveyor belt speed; The number of computer motherboards soldered per unit time is used as an efficiency indicator. The energy consumption index is calculated by dividing the energy consumption per unit time by the number of computer motherboards soldered per unit time. Welding quality is evaluated by weighted comprehensive assessment of solder joint yield and component offset, which serves as the quality indicator. Based on efficiency, energy consumption, and quality indicators, the TOPSIS method is used to select the optimal [result]. and the optimal The corresponding optimization weight coefficients are used as the final determined optimization weight coefficients.

6. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 1, characterized in that, The process of acquiring image information of the computer motherboard to be soldered, identifying the component layout on the motherboard, and dividing the thermal management area based on the thermal characteristics of the components specifically includes: The motherboard topology image is acquired using a high-resolution camera, and then noise reduction and contrast enhancement are performed. Deep learning object detection algorithms are used to identify various components in images and record their type, package size, and precise coordinates on the motherboard. Based on the known thermal characteristics of the components, the motherboard surface is divided into a high heat capacity area, a heat-sensitive area, and a standard soldering area, and a corresponding thermal management label is generated for each area.

7. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 1, characterized in that, The process of adjusting the optimal welding temperature curve based on the defined heat management regions, and generating a fine-tuning strategy for temperature zone parameters for different heat management regions, specifically includes: Based on the geometric model of the motherboard and the division of the thermal management area, thermal field simulation is performed to predict the actual temperature of different thermal management areas during the reflow process. Compare the simulation results with the target temperature curve, calculate the required increase in temperature compensation for high heat capacity regions, and calculate the required decrease in temperature and / or slowing down the heating rate for heat-sensitive regions. The calculated compensation values ​​for each region are integrated into the target temperature curve to generate a final temperature control curve with spatially differentiated parameters.

8. The method for correcting and controlling the soldering temperature of a computer motherboard based on machine learning according to claim 1, characterized in that, The process of synthesizing a final temperature control instruction set based on the optimal welding temperature curve and the fine-tuning strategy to control the reflow oven for mass production specifically includes: During mass production, thermocouples and vision systems are used to collect real-time images of the actual temperature curve inside the furnace and images of key welding points. The real-time data is compared with the expected final temperature curve. If the deviation continues to exceed the tolerance range, an adjustment decision is triggered. Based on the decision results, the corresponding temperature zone setpoint of the reflow oven is automatically and gradually corrected to restore the process to a controlled state.

9. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a machine learning-based computer motherboard soldering temperature correction control method as described in any one of claims 1-8.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, they implement the machine learning-based computer motherboard soldering temperature correction control method according to any one of claims 1-8.

Images