A MOSFET Collinear Packaging Control Method and System

Abstract

The present invention discloses a MOSFET collinear packaging control method and system, which relates to the technical field of semiconductor manufacturing. The method includes: obtaining a chip distribution frame diagram of a wafer to be cut, and performing optimization of the shortest cutting path according to the diagram to generate a recommended wafer cutting path; obtaining timing information of cutting speed and depth. Through a cutting temperature prediction model, combining cutting path, speed and depth information, cutting temperature timing information is generated; optimizing cooling control according to the temperature timing information to obtain recommended coolant flow rate and pressure timing information; performing wafer cutting based on cutting and cooling parameters, and obtaining multiple separated chips for MOSFET collinear integrated packaging. It solves the technical problems of inaccurate temperature control and low cooling efficiency in the prior art during wafer cutting, and achieves the technical effect of improving the accuracy of temperature control and cooling efficiency during wafer cutting.

Description

Technical Field

[0001] This application relates to the field of semiconductor manufacturing technology, and particularly to a method and system for controlling MOSFET collinear packaging. Background Art

[0002] In the production process of modern semiconductor devices, wafer dicing is a crucial process step. Especially in the manufacturing of MOSFET collinear packaging, the dicing accuracy directly affects the chip performance and packaging quality. The temperature control methods in the prior art usually rely on real-time monitoring and adjustment, which have the problem of long delay, resulting in inaccurate temperature control and low cooling efficiency. This not only affects the stability of the dicing process but also increases the equipment loss and manufacturing cost. During the dicing process, the heat generated by the changes in dicing speed and depth may cause local overheating, affecting the dicing accuracy, especially in applications that require high-precision packaging. How to accurately control the temperature and optimize the cooling process to improve the dicing accuracy and stability has become a difficult problem in the current technology.

[0003] In the current related technologies, there are technical problems of inaccurate temperature control and low cooling efficiency in the wafer dicing process. Summary of the Invention

[0004] This application solves the technical problems of inaccurate temperature control and low cooling efficiency in the wafer dicing process in the prior art by providing a method and system for controlling MOSFET collinear packaging.

[0005] This application provides a method for controlling MOSFET collinear packaging, including:

[0006] Obtaining a chip distribution frame diagram of the wafer to be diced; optimizing the shortest dicing path according to the chip distribution frame diagram to obtain a recommended wafer dicing path; obtaining dicing speed timing information and dicing depth timing information; processing the recommended wafer dicing path, the dicing speed timing information, and the dicing depth timing information through a dicing temperature prediction model to obtain dicing temperature timing information; optimizing the cooling control according to the dicing temperature timing information to obtain recommended coolant flow rate timing information and recommended coolant pressure timing information; performing wafer dicing according to the dicing speed timing information, the dicing depth timing information, the recommended coolant flow rate timing information, and the recommended coolant pressure timing information to obtain a plurality of separated chips for MOSFET collinear integrated packaging.

[0007] This application provides a system for controlling MOSFET collinear packaging, including:

[0008] Chip distribution frame diagram acquisition module, which is used to obtain the chip distribution frame diagram of the wafer to be cut; cutting shortest path optimization module, which is used to optimize the cutting shortest path according to the chip distribution frame diagram to obtain the recommended wafer cutting path; timing information acquisition module, which is used to obtain the cutting speed timing information and the cutting depth timing information; cutting temperature timing information acquisition module, which is used to process the recommended wafer cutting path, the cutting speed timing information and the cutting depth timing information through a cutting temperature prediction model to obtain the cutting temperature timing information; cooling control optimization module, which is used to optimize the cooling control according to the cutting temperature timing information to obtain the recommended coolant flow rate timing information and the recommended coolant pressure timing information; wafer cutting module, which is used to cut the wafer according to the cutting speed timing information, the cutting depth timing information, the recommended coolant flow rate timing information and the recommended coolant pressure timing information to obtain a plurality of separated chips for MOSFET collinear integrated packaging.

[0009] It is proposed to use a MOSFET collinear packaging control method and system in this application. First, obtain the chip distribution frame diagram of the wafer to be cut, optimize the cutting shortest path according to this diagram to generate the recommended wafer cutting path, and obtain the timing information of the cutting speed and depth. Through a cutting temperature prediction model, combine the cutting path, speed and depth information to generate the cutting temperature timing information; optimize the cooling control according to the temperature timing information to obtain the recommended coolant flow rate and pressure timing information; perform wafer cutting based on the cutting and cooling parameters to obtain multiple separated chips for MOSFET collinear integrated packaging, achieving the technical effect of improving the accuracy of temperature control and the cooling efficiency during wafer cutting. Brief Description of the Drawings

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments of the present invention will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of this application. It should be understood that the operations in the front or below do not necessarily need to be executed precisely in sequence. On the contrary, according to the need, they can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0011] Figure 1 It is a schematic flowchart of a MOSFET collinear packaging control method provided by an embodiment of this application;

[0012] Figure 2It is a schematic structural diagram of a MOSFET collinear packaging control system provided by an embodiment of the present application.

[0013] Explanation of reference numerals: Chip distribution frame diagram acquisition module 10, cutting shortest path optimization module 20, timing information acquisition module 30, cutting temperature timing information acquisition module 40, cooling control optimization module 50, wafer cutting module 60. Detailed implementation manners

[0014] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically gives the detailed implementation manners of the present application.

[0015] In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0016] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first\second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0017] An embodiment of the present application provides a MOSFET collinear packaging control method, as Figure 1 shown, the method includes:

[0018] Step S100: Obtain the chip distribution frame diagram of the wafer to be cut. Specifically, select an optical imaging device with high resolution and a resolution of micron level or higher, such as an industrial microscope or a wafer inspection optical scanning system. Debug and calibrate the lens focal length, light source brightness angle, set an appropriate magnification, and prepare the data acquisition and storage module. Then place the wafer at the center of the imaging device stage and perpendicular to the optical axis, record the initial position using positioning marks, and you can choose to image from different angles and multiple perspectives. Control the imaging device to scan point by point according to the preset method, and transmit and save the image data in real time. Then preprocess the collected original images, use algorithms such as median filtering and Gaussian filtering to reduce noise, and use histogram equalization or contrast stretching techniques to enhance the contrast. Then apply an edge detection algorithm such as the Canny operator, and accurately extract the chip edge information through steps such as Gaussian filtering, gradient calculation, non-maximum suppression, and double-threshold detection. Then use mathematical morphological closing operation to connect the edges, and use the least squares method to fit the curve to construct a chip distribution frame diagram represented by a polygon or a rectangle, clearly showing the chip position, shape, and layout, providing basic data for the subsequent cutting process.

[0019] Step S200: Optimize the shortest cutting path according to the chip distribution frame diagram to obtain the recommended wafer cutting path. Specifically, perform digital processing on the chip distribution frame diagram, extract and sort out the key feature information of the chips, and select a suitable path optimization algorithm model according to the wafer cutting requirements and constraints, such as Dijkstra, A* algorithm, or genetic and simulated annealing algorithms, etc. Then construct a graph structure with the chip boundaries and connection points as graph nodes and the cutting path as edges, and assign weights to the edges according to factors such as cutting distance, difficulty, chip importance, and thermal effect. Select the starting and target nodes. For graph theory algorithms, start expanding adjacent nodes from the starting node and update the path cost and predecessor node information until the target node is taken out; heuristic algorithms randomly generate an initial cutting path solution as a population, evaluate it according to the fitness function, and iteratively optimize it through genetic operations or annealing strategies. Then optimize and verify the preliminary shortest path, check whether it meets the constraints such as the physical limitations of the cutting equipment, process requirements, and conflict avoidance, adjust and correct it through local search, backtracking method, etc., and at the same time analyze the actual uncertainty factors for redundant design and fault tolerance analysis. The finally determined recommended wafer cutting path can shorten the cutting stroke, ensure the chip quality, provide a basis for the co-linear packaging of MOSFETs, and improve production efficiency and economic benefits.

[0020] Step S300, obtain the cutting speed timing information and cutting depth timing information. Specifically, the cutting equipment and sensors are selected and installed. For the cutting speed, a high-precision laser Doppler speed meter or encoder is selected, installed on the moving parts of the cutting tool and calibrated; for the cutting depth, a high-precision LVDT or capacitive displacement sensor is used, installed on the tool holder and other components and calibrated. Then, a data acquisition system is built, including a data acquisition card, a signal conditioning circuit and a computer software. The speed and depth data acquisition are synchronized through a unified clock or timestamp, and the sensor signals are collected at a high sampling frequency and converted and stored. Then, the sensor is preheated and initialized before the cutting equipment is started. During cutting, data is continuously collected and recorded in real time. At the same time, the data is preliminarily analyzed. If there is an abnormal fluctuation, the equipment and sensor are checked to ensure that the cutting is normal and the data is accurate. Finally, the collected data is post-processed, and a filtering algorithm is used to remove noise, check integrity and interpolate to repair missing data, compare the preset process parameters to verify accuracy, find out the cause of deviation and correct it, and provide a reliable data basis for subsequent process optimization.

[0021] Step S400, through the cutting temperature prediction model, the wafer cutting recommended path, the cutting speed timing information and the cutting depth timing information are processed to obtain the cutting temperature timing information. Specifically, a large amount of historical wafer cutting data covering different wafer materials, cutting equipment parameters, paths and corresponding speeds, depth settings and actual cutting temperature monitoring values ​​are collected, pre-processed, checked for integrity and accuracy, abnormal data are removed, and interpolation and supplementation of missing values ​​are normalized. Then, according to the characteristics of wafer cutting, a suitable prediction model architecture is selected, such as a heat conduction model based on physical principles, an empirical formula model or a machine learning model (taking a neural network as an example), and a multi-layer perceptron structure including an input layer (receiving characteristic parameters of the wafer cutting recommended path, cutting speed and depth timing information), a hidden layer (determining the number of layers and nodes through experiments, and selecting a nonlinear activation function) and an output layer (outputting the predicted temperature value) is constructed. The pre-processed data is divided into a training set, a validation set and a test set to train the model, and a suitable loss function and optimization algorithm are used to iterate the training until convergence to obtain a cutting temperature prediction model. Finally, the model is verified with an independent test set, and the predicted temperature timing information is compared with the actual measured value. The error index is calculated to evaluate the accuracy. If it is not ideal, it is optimized by adjusting the model structure parameters, training algorithm parameters, or increasing the diversity of training data to enable accurate prediction, provide a reliable basis for cooling control and cutting process optimization, and ensure chip quality and manufacturing process stability.

[0022] In a possible implementation, the wafer cutting recommended path, the cutting speed timing information, and the cutting depth timing information are processed through a cutting temperature prediction model to obtain cutting temperature timing information. Step S400 further includes step S410 of obtaining wafer cutting equipment parameters, where the wafer cutting equipment parameters include cutting channel structure parameters, cutting groove structure parameters, connection rib structure parameters, and cutting machine precision parameters. Specifically, to obtain the wafer cutting equipment parameters, high-precision measuring instruments are required for measurement and evaluation operations. For the cutting channel structure parameters, an electron microscope or a laser profiler is used to accurately measure its width to the micron level. Since the width affects heat dissipation and material removal rate, the depth is also measured as it is related to the tool penetration depth and affects energy and temperature distribution. At the same time, the surface roughness is recorded. These parameters are of great significance for building the model. For the cutting groove structure parameters, precision equipment is used to determine its shape, size, and the inclination angle of the groove wall. Different shapes and sizes will have a significant impact on the cutting stress distribution and the coolant flow characteristics, providing detailed information for temperature prediction. For the connection rib structure parameters, its thickness, height, and width are measured. The thickness affects the cutting resistance and heat, and the proportional relationship is related to the stress and heat conduction path. The material composition and thermal conductivity also need to be determined to help understand the heat propagation law. When evaluating the cutting machine precision parameters, a standard calibration block and a coordinate measuring machine are used to measure the positioning errors of the tool in the X, Y, and Z directions. These errors affect the accuracy of the cutting path and heat generation. The motion repeatability is measured to judge the cutting stability and temperature fluctuation. The feed speed control precision is analyzed to ensure cutting according to the preset speed timing and avoid excessive heat generation. The precision parameters enable us to fully analyze the impact of uncertain factors during cutting on the temperature.

[0023] Step S420, the cutting temperature prediction model is constructed with the cutting path structural parameters, the cutting groove structural parameters, the connecting rib structural parameters, and the cutting machine precision parameters as constraints. Specifically, the model selection and architecture design are carried out based on the physical characteristics of the wafer cutting process and the characteristics of the acquired equipment parameters, and a finite element analysis model (FEM) or a neural network model can be selected. The finite element analysis model discretizes the cutting process into small units based on the principles of heat conduction, convection and radiation to solve the heat balance equation to predict the temperature distribution, but the calculation is complex and requires in-depth analysis of the physical process and precise setting of parameters; the neural network model has strong nonlinear fitting capabilities and is suitable for processing complex multivariate problems. Taking it as an example, a multilayer perceptron architecture with input, hidden, and output layers is designed. The input layer nodes are determined according to parameters such as cutting paths, grooves, connecting ribs, and cutting machine precision. The hidden layer first tries 2 to 3 layers, with dozens to hundreds of nodes in each layer, and uses activation functions such as ReLU and tanh to enhance the expressiveness. The output layer outputs the predicted temperature value. When constructing the model, the cutting paths, grooves, connecting rib structural parameters, and cutting machine precision parameters are incorporated as constraints. In the finite element model, parameters are used to define geometric shapes, material properties and boundary conditions, such as setting the heat conduction boundary according to the width and depth of the cutting path, determining the coolant flow channel and heat transfer coefficient according to the cutting groove, setting the conduction coefficient according to the thermal conductivity of the connecting ribs, and considering the impact of the uncertainty of the cutting path on the temperature according to the accuracy of the cutting machine; the neural network model adds these parameters as part of the feature vector in the training data to automatically capture the relationship, and uses L1 or L2 regularization to prevent overfitting, ensuring that the model accurately predicts the temperature based on the constraints, and improves generalization and prediction accuracy. The model is trained using a large amount of experimental or simulated data, with the actual measured temperature as the target value, and the model parameters are continuously adjusted, such as the material property parameters of the finite element, the weights and biases of the neural network, to minimize the error between the predicted value and the actual value. After multiple iterations to the preset convergence conditions, such as the mean square error is less than the threshold or reaches the predetermined number of training rounds, the obtained model can reflect the impact of equipment parameters on temperature and provide a reliable tool for temperature prediction.

[0024] Step S430: Process the recommended wafer cutting path, the cutting speed timing information, and the cutting depth timing information through the cutting temperature prediction model to obtain the cutting temperature timing information. Specifically, before applying the cutting temperature prediction model, relevant data needs to be prepared and preprocessed. For the recommended wafer cutting path, convert it into geometric feature parameters such as path length, number of turning points, and path curvature to more intuitively show the complexity of the cutting path and its potential impact on temperature; perform normalization operations on the cutting speed and cutting depth timing information to map their numerical ranges to the interval [0,1] to ensure comparability with other parameter orders of magnitude. At the same time, check the integrity and continuity of the data. If there are missing data or outliers, interpolation or correction methods should be used in a timely manner to ensure the reliability of the data input into the model. After the preprocessing is completed, the wafer cutting recommended path feature parameters, the cutting speed and depth timing information, and the previously obtained wafer cutting equipment parameters (including cutting lanes, cutting grooves, joint rib structure parameters, and cutting machine accuracy parameters) are used as inputs and fed into the constructed cutting temperature prediction model. The model performs a series of computational processes on the input data based on its internal mathematical model and trained parameters. The finite element analysis model obtains the temperature distribution at each time step during the cutting process by solving the heat conduction equation, and the neural network model calculates the predicted temperature value at each time step using the forward propagation algorithm, thereby generating the cutting temperature timing information. During the prediction process, closely monitor the model output results to check for unreasonable situations such as too high or too low temperature, temperature mutations, etc. Once an anomaly is found, immediately check the accuracy of the input data and the model running status, eliminate possible problems, ensure that the obtained cutting temperature timing information is accurate and error-free, truly reflect the temperature change trend during wafer cutting, provide strong support for subsequent cooling control optimization and wafer cutting process optimization, promote the smooth progress of subsequent processes such as MOSFET co-packaging, improve the quality and efficiency of the semiconductor manufacturing process, and reduce the product defect rate and production cost caused by temperature problems.

[0025] In a possible implementation manner, a cutting temperature prediction model is constructed with the cutting channel structure parameters, the cutting groove structure parameters, the connecting rib structure parameters, and the cutting machine precision parameters as constraints. Step S420 further includes step S421 of configuring a cutting channel structure similarity threshold, a cutting groove structure similarity threshold, a connecting rib structure similarity threshold, and a cutting machine precision deviation distance threshold. Specifically, various thresholds are determined based on in-depth analysis of the wafer cutting process and previous empirical data. For the cutting channel structure similarity threshold, considering the important influence of its width, depth, and surface roughness on the cutting temperature, after analyzing the temperature change laws under different combinations and the temperature control precision requirements of actual production, for example, when the cutting channel width deviation is within ±5 μm, the depth deviation is within ±10 μm, and the surface roughness deviation is within ±0.5 μm, the influence on the temperature is acceptable, and the deviation range is converted into a comprehensive threshold for sample screening. The cutting groove structure similarity threshold is set according to the influence mechanism of its shape, dimensions (groove width, groove depth, groove length, and spacing), and groove wall inclination angle on the temperature. By simulating the coolant flow and heat transfer conditions, it is determined that when the groove width and groove depth deviations are within ±15%, and the groove wall inclination angle deviation is within ±5 degrees (same shape), the influence on the temperature is relatively small, so as to ensure the similarity of the sample cutting groove structure and the reliability of the temperature prediction basis. When determining the connecting rib structure similarity threshold, considering the close relationship between its thickness, height, width, and material thermal conductivity and the cutting stress distribution and heat conduction path, through mechanical and thermal analysis and sensitivity tests, a similarity judgment criterion is set with the thickness deviation within ±8%, the height and width deviations within ±10%, and the material thermal conductivity deviation within ±15%, to ensure the comparability and regularity of the influence of the sample connecting rib structure on the temperature. When setting the cutting machine precision deviation distance threshold, for key indicators such as the positioning precision, motion repeatability, and feed speed control precision of the cutting tool, according to the equipment technical specifications and actual cutting precision requirements, for example, the positioning precision is set to ±0.02 mm according to the equipment nominal value of ±0.01 mm combined with the actual fluctuation, the motion repeatability is set with a certain multiple (such as 2 times) of the standard deviation of the tool position deviation in multiple repeated cutting experiments as the threshold, and the feed speed control precision is set to a speed deviation percentage of ±5% according to the speed stability requirements and the influence of speed fluctuation on the temperature as the threshold, so as to ensure that the samples included in the model have similar cutting machine precision conditions and improve the accuracy and reliability of the model's prediction of the cutting temperature.

[0026] Step S422: Obtain the first cutting sample, where the first cutting sample has a cutting channel structure label, a cutting groove structure label, a connecting rib structure label, and a cutting machine accuracy label. Specifically, establish a sample collection system covering various situations, and record in detail the wafer cutting process under different models and process parameter settings. During each cutting experiment, accurately measure the width, depth, and surface roughness of the cutting channel and organize them into a cutting channel structure label; accurately record the shape, size, and groove wall inclination angle of the cutting groove to generate a cutting groove structure label; measure the thickness, height, width, and material thermal conductivity of the connecting rib to form a connecting rib structure label; use high-precision instruments such as a coordinate measuring machine and a laser interferometer to measure the positioning accuracy, motion repeatability, and feed speed control accuracy of the cutting tool, and convert them into a cutting machine accuracy label. Thus, a first cutting sample with complete labels is collected, providing data resources for subsequent work. During the collection process, ensure the diversity and representativeness of the samples, including different wafer materials (such as common silicon wafers and wafers with special doping and composite structures), different cutting process parameters (set a cutting speed range from low to high and a cutting depth range from shallow to deep and conduct multiple experiments), and different cutting equipment (select equipment from different manufacturers, with different accuracy levels and technical characteristics), so that the first cutting sample can reflect various actual production situations, enhancing the generalization ability and prediction accuracy of the model.

[0027] Step S423: Compare the cutting track structure label, the cutting groove structure label, the connecting rib structure label, and the cutting machine precision label with the cutting track structure parameters, the cutting groove structure parameters, the connecting rib structure parameters, and the cutting machine precision parameters to obtain the cutting track structure similarity, the cutting groove structure similarity, the connecting rib structure similarity, and the cutting machine precision deviation distance. Specifically, perform sample comparison calculations, which are based on a preset similarity calculation method to accurately compare and calculate the similarity between the structure labels and precision labels of the first cutting sample and the cutting track, groove, connecting rib structure parameters, and cutting machine precision parameters of the model to be constructed. To calculate the cutting track structure similarity, use a weighted distance algorithm that comprehensively considers width, depth, and surface roughness. For example, assign weights of 0.4, 0.4, and 0.2 to these three items respectively, calculate the difference between the sample and the target parameters in each dimension and sum them after weighting to obtain an index that takes values between 0 and 1. The closer the index is to 1, the higher the similarity. To calculate the cutting groove structure similarity, for its shape, size, and groove wall inclination angle, first convert the shape into numerical encoding, then combine parameters such as size into a feature vector, and calculate it with the target vector using the cosine similarity algorithm. The above calculation method is accurate and stable for complex factors. To calculate the connecting rib structure similarity, use the relative deviation method according to its parameter characteristics, calculate the deviation of each sample parameter relative to the target respectively and perform weighted averaging (such as weighting according to 0.3, 0.2, 0.2, 0.3) to obtain an index, which reflects the degree of closeness to the target structure parameters to judge the similarity of the influence of temperature. To calculate the cutting machine precision deviation distance, directly measure the difference between the sample and the target precision parameters and perform weighted summation according to the preset weights (positioning precision 0.5, motion repeatability 0.3, feed speed control precision 0.2) to obtain an index, which intuitively reflects the degree of deviation, provides a quantitative basis for sample screening, ensures the consistency of the sample and the target cutting conditions in terms of precision, and reduces the model prediction error caused by precision differences.

[0028] Step S424, when the similarity of the cutting channel structure is greater than or equal to the cutting channel structure similarity threshold, the similarity of the cutting groove structure is greater than or equal to the cutting groove structure similarity threshold, the similarity of the connecting rib structure is greater than or equal to the connecting rib structure similarity threshold, and the precision deviation distance of the cutting machine is less than or equal to the precision deviation distance threshold of the cutting machine, add the first cutting sample into the sample set for constructing the cutting temperature prediction model. Specifically, establish an automated sample screening process. After calculating the similarities of the cutting channel, cutting groove, and connecting rib structures and the precision deviation distance of the cutting machine, compare the results with the preset thresholds one by one. If the similarity of the cutting channel structure of a certain first cutting sample meets the standard, it indicates that the key parameters of its cutting channel structure for temperature are similar to the target cutting conditions and can provide temperature prediction information for the model; if the similarity of the cutting groove structure meets the requirements, the cutting groove structure characteristics of the sample are similar to the target, and the influence of coolant flow and heat transfer characteristics on temperature can be compared; if the similarity of the connecting rib structure meets the standard, it shows that its influence law on temperature is similar to the target and helps the model to learn; if the precision deviation distance of the cutting machine is within the threshold, it ensures that the precision of the sample and the target cutting process is consistent, reducing the interference of precision differences on model prediction. Only when the sample meets all these conditions can it be included in the sample set for constructing the cutting temperature prediction model and used as the basic data for subsequent model training. Through strict screening, ensure that the samples have high similarity and representativeness, accurately reflect the temperature change law under the target cutting conditions, and improve the model training and prediction effects. When adding samples, record the number of each sample in detail, covering the original cutting experiment conditions, various parameters, and the calculation results of similarities and deviation distances, which is convenient for retrospective analysis in model training optimization, understanding the learning response of the model to sample characteristics, and helping to improve and perfect the model.

[0029] Step S425: When the sample set for constructing the cutting temperature prediction model is greater than or equal to the preset sample quantity, train the cutting temperature prediction model. Specifically, a reasonable preset sample quantity is set, which requires comprehensive analysis of factors such as model complexity, data diversity, and requirements for prediction accuracy. More complex models (such as neural network models with many hidden layers and nodes) require more samples to avoid overfitting; if the samples cover various wafer materials, cutting process parameters, and equipment types, relatively fewer samples may also meet the basic training requirements. For example, for a neural network cutting temperature prediction model of medium complexity, if the samples cover more than 3 types of wafer materials, more than 5 combinations of process parameters, and more than 2 types of equipment, the preset sample quantity can be set to about 500; if the model complexity is low or the data diversity is small, it can be appropriately reduced, but generally not less than 200, to ensure that the model learns sufficient temperature change laws and characteristic relationships to have the prediction ability. When the sample quantity in the sample set for constructing the cutting temperature prediction model reaches the standard, start the training program. Select appropriate training algorithms and parameter settings according to the selected model type (finite element analysis model, neural network model, etc.). For the finite element model, determine physical parameters such as the thermal conductivity coefficient and boundary conditions, as well as the solution algorithm and the number of iterations; for the neural network model, select activation functions (such as ReLU, tanh, etc.), optimization algorithms (such as stochastic gradient descent, Adam optimizer, etc.), and learning rate, regularization parameters, etc., and then divide the sample set into a training set, a validation set, and a test set according to the ratio of 80%, 10%, and 10% to start training. During training, continuously adjust the model parameters to gradually reduce the error between the prediction result of the training set and the actual temperature value, and use the validation set to monitor the performance to prevent overfitting. After multiple iterations until the performance indicators of the validation set (such as mean square error, mean absolute error, etc.) reach a stable and satisfactory level, the model training is completed, and it can accurately predict the temperature under the target cutting conditions and be applied to the actual wafer cutting temperature prediction and control, helping to improve the quality and efficiency of semiconductor manufacturing.

[0030] Step S500: Optimize the cooling control according to the cutting temperature timing information to obtain the coolant recommended flow rate timing information and the coolant recommended pressure timing information. Specifically, data related to the cutting process need to be collected, including the cutting track, groove, connection rib structure parameters, cutting machine precision parameters, and accurate cutting temperature timing information. At the same time, high-precision temperature and ambient temperature monitoring equipment are used to ensure the accuracy of the data, laying a foundation for optimizing the cooling control. Then, cooling sample collection experiments are carried out with various parameters and temperature information as constraints. The wafer is cut under different combinations of coolant flow rate and pressure, and the temperature, flow rate, and pressure data are recorded to form a sample set. When the number of samples exceeds the threshold, a two-dimensional virtual coordinate system is constructed with the normalized data of the coolant flow rate and pressure and the samples are mapped. Then, LOF outlier analysis is performed on the samples in the coordinate system to exclude outliers, and the concentrated cooling samples with an outlier factor less than or equal to the threshold are extracted. Mean analysis is performed on them to obtain the coolant recommended flow rate timing information and the coolant recommended pressure timing information, providing accurate parameters for the cooling control system, ensuring the cutting quality and subsequent processes, and improving the performance and reliability of semiconductor manufacturing.

[0031] In a possible implementation manner, to optimize the cooling control according to the cutting temperature timing information to obtain the coolant recommended flow rate timing information and the coolant recommended pressure timing information, step S500 further includes step S510: obtain the ambient temperature monitoring value. Specifically, a plurality of high-precision temperature sensors are arranged near the wafer cutting working area. The sensors should have the capabilities of fast response and accurate measurement, and can monitor the change of the ambient temperature in real time. The selection of the sensor position is crucial and it is necessary to avoid being directly affected by factors such as the heat generated by the cutting equipment itself and the heat dissipation of the coolant, so as to ensure that the collected temperature data can truly reflect the ambient temperature condition of the cutting area. The data acquisition system is used to regularly (for example, every few seconds) record and summarize the measurement values of the sensors to form a continuous sequence of ambient temperature monitoring values, providing basic data support for subsequent cooling control analysis. Calibrate and maintain the sensors to ensure that their measurement accuracy remains stable throughout the wafer cutting process, and avoid inaccurate ambient temperature monitoring data caused by sensor failures or errors, thus affecting the accuracy of cooling sample collection and coolant parameter optimization.

[0032] Step S520: Obtain the wafer dicing equipment parameters. Among them, the wafer dicing equipment parameters include dicing lane structure parameters, dicing groove structure parameters, connection rib structure parameters, and dicing machine precision parameters. Specifically, to obtain comprehensive and accurate wafer dicing equipment parameters, a variety of high-precision measurement tools and technologies are required. For the dicing lane structure parameters, an optical measuring instrument or an electron microscope is used to measure the dicing lane width at multiple positions and take the average value to achieve micron-level accuracy. Combining a depth measurement probe with image analysis technology to determine the dicing lane depth, and using a surface roughness measuring instrument to record its surface roughness. These parameters affect the flow resistance of the coolant, the heat dissipation area, the energy consumption, and the temperature distribution, and are all recorded in detail for subsequent reference. For the dicing groove structure parameters, three-dimensional laser scanning technology or precision mechanical measuring tools are used to accurately measure its shape (such as rectangular, V-shaped, U-shaped, etc.), dimensions (groove width, groove depth, length, and spacing, etc., strictly controlling the measurement error), and the groove wall inclination angle (obtained through an angle measuring instrument or an image algorithm). Since different shapes and sizes of dicing grooves have a significant impact on the flow and heat transfer of the coolant, the measured parameters will be sorted out for subsequent work. For the connection rib structure parameters, a caliper and a micrometer are used to measure the thickness, and a three-dimensional coordinate measuring instrument is used to measure the height and width to clarify the influence of its proportional relationship on the wafer stress distribution and the heat conduction path. A thermal conductivity measuring instrument or a reference manual is used to determine the material composition and thermal conductivity. The measured parameters are accurately recorded and incorporated into the system. In terms of evaluating the dicing machine precision parameters, a standard calibration block and a coordinate measuring machine are used to measure the cutting tool positioning accuracy in the X, Y, and Z directions multiple times and calculate the standard deviation. Repeatedly cut the same path to measure the position deviation to statistically analyze the motion repeatability. Use a speed sensor and the feedback data of the control system to evaluate the feed speed control accuracy. These precision parameters are recorded and analyzed in detail, serving as important constraints for cooling sample collection and coolant parameter optimization, ensuring that the coolant flow rate and pressure adapt to the actual precision of the dicing machine and achieving effective cooling control.

[0033] Step S530: Cooling sample collection is carried out with the cutting channel structure parameters, the cutting groove structure parameters, the connecting rib structure parameters, the cutting machine precision parameters, and the cutting temperature time series information as constraints to obtain a cooling sample set. Among them, any one cooling sample has coolant flow rate recording time series data and coolant pressure recording time series data. Specifically, build a cooling sample collection experimental platform that can accurately control the flow rate and pressure of the coolant and monitor the temperature changes and other relevant parameters during the cutting process in real time. According to the pre-determined cutting channel, cutting groove, and connecting rib structure parameters, as well as the cutting machine precision parameters, set the initial state of the cutting equipment to ensure that the cutting conditions are consistent for each experiment. Conduct wafer cutting experiments under different combinations of coolant flow rate and pressure. The value ranges of the flow rate and pressure should be reasonably determined based on actual production experience and theoretical analysis. For example, the coolant flow rate can gradually increase from a lower starting value to a higher upper limit value, and the pressure also changes within a certain range accordingly, forming multiple different flow rate-pressure combination working conditions. Under each working condition, use a data acquisition system to record the changes in the coolant flow rate and pressure over time during the cutting process, forming the coolant flow rate recording time series data and the coolant pressure recording time series data. At the same time, record the cutting temperature time series information and other relevant process parameters (such as cutting force, chip morphology, etc. These parameters can be used as auxiliary information for in-depth analysis of the cooling effect in the future). Organize the data such as the coolant flow rate, pressure, and cutting temperature obtained from each experiment into a cooling sample. After multiple experiments under different working conditions, accumulate a large number of cooling samples to form a cooling sample set. These samples reflect the combination of specific cutting equipment parameters and cutting temperature changes under different cooling conditions, providing a rich data basis for subsequent optimization of coolant parameters. By analyzing and mining these samples, the most suitable combination of coolant flow rate and pressure can be found to effectively control the cutting temperature and improve the quality and efficiency of wafer cutting.

[0034] Step S540: When the number of samples in the cooling sample set is greater than the cooling sample number threshold, based on the chronological data of the coolant flow rate record and the chronological data of the coolant pressure record, distribute the cooling sample set in a two-dimensional virtual coordinate system. Specifically, determine a cooling sample number threshold. The setting of this threshold requires comprehensive analysis of factors such as the complexity of the model, the diversity of data, and the accuracy requirements for optimizing coolant parameters. For a relatively complex cooling model (such as a high-precision model considering the interaction of multiple factors), more sample numbers are needed to fully capture the complex relationship between the coolant flow rate, pressure, and cutting temperature. For a relatively simple model, the sample number can be appropriately reduced, but it is also necessary to ensure that it can cover a sufficient range of cooling operating conditions to ensure the reliability of the optimization result. Usually, a combined method of experiments and simulation analysis can be used to evaluate the model performance under different sample numbers, so as to determine a suitable cooling sample number threshold. When the number of samples in the cooling sample set reaches or exceeds this threshold, start processing the samples for constructing a two-dimensional virtual coordinate system. Use the normalized value of the chronological data of the coolant flow rate record as the abscissa, and normalize the flow rate data of each cooling sample so that its numerical range is between [0, 1] for unified comparison and analysis in the coordinate system. Use the normalized value of the chronological data of the coolant pressure record as the ordinate, and similarly normalize the pressure data to construct a two-dimensional virtual coordinate system. Map each sample in the cooling sample set to the two-dimensional virtual coordinate system according to its corresponding normalized flow rate and pressure values, so that each sample has a corresponding position in the coordinate system. Through the above method, the originally complex cooling sample data is intuitively displayed on the two-dimensional plane, facilitating subsequent central point sorting and coolant parameter optimization operations, and enabling a clearer observation of the distribution law and trend of the samples, and finding potential optimal combination regions of the coolant flow rate and pressure.

[0035] Step S550: Perform center point sorting based on the two-dimensional virtual coordinate system to obtain the coolant recommended flow rate time series information and the coolant recommended pressure time series information. Specifically, in the constructed two-dimensional virtual coordinate system, perform center point sorting operations on the cooling sample points to determine the coolant recommended flow rate time series information and the coolant recommended pressure time series information. Use a clustering analysis algorithm (such as the K-Means clustering algorithm) to cluster the cooling sample points in the coordinate system, clustering the sample points with similar flow rate and pressure characteristics into one category. Through multiple iterative calculations, determine several main clustering centers. Perform density analysis on the sample points around each clustering center, calculate its local density, for example, use a distance-based density calculation method to determine the sample density of each clustering center. According to the sample density of the clustering center and its relative distance from other clustering centers, calculate the comprehensive score of each clustering center. A clustering center with a higher comprehensive score indicates that the sample points around it are more concentrated and representative, and are more likely to contain the optimal coolant flow rate and pressure combination. Select the clustering center with the highest comprehensive score as the final center point. The normalized values of the coolant flow rate and pressure corresponding to this center point, after inverse normalization processing, obtain the actual coolant recommended flow rate time series information and the coolant recommended pressure time series information. The recommended information reflects the variation of the optimal coolant flow rate and pressure that can effectively control the temperature during the cutting process under the constraints of the given cutting equipment parameters and cutting temperature time series information, providing precise control parameters for the actual wafer cutting cooling control system. It can dynamically adjust the flow rate and pressure of the coolant according to the real-time situation of the cutting process, ensure that the cutting temperature is maintained within an appropriate range, improve the quality and stability of wafer cutting, reduce chip defects and material damage caused by temperature problems, provide a high-quality wafer foundation for subsequent semiconductor manufacturing processes, and enhance the production efficiency and economic benefits of the entire semiconductor manufacturing process.

[0036] In a possible implementation, when the number of samples in the cooling sample set is greater than the cooling sample number threshold, according to the coolant flow rate recorded time series data and the coolant pressure recorded time series data, distribute the cooling sample set in a two-dimensional virtual coordinate system. Step S540 further includes step S541, using the normalized data of the coolant flow rate to construct the first coordinate axis. Specifically, in the data preparation stage, first collect the coolant flow rate recorded time series data in each cooling sample obtained from the cooling sample collection experiment, which reflects the actual changes in the coolant flow rate under different cutting conditions and time nodes. Then perform normalization processing on the flow rate data. Adopt the linear normalization method. For example, if the minimum and maximum values of the original flow rate data are known, convert the flow rate value at a certain moment into a normalized value within the range of [0, 1], so that all flow rate data have the same order of magnitude and value range, which is convenient for subsequent intuitive comparison and analysis. In the coordinate axis construction step, select a two-dimensional coordinate environment supported by computer graphics or data analysis software as the drawing plane. Use the normalized coolant flow rate data value as the abscissa and draw the first coordinate axis along the horizontal direction. Reasonably determine the scale interval according to the data accuracy and distribution (such as 0.1 or 0.05, etc.) to clearly locate the corresponding points of the cooling samples, complete the construction of the first coordinate axis, and lay the foundation for the subsequent configuration of the two-dimensional virtual coordinate system to display the changes in the coolant flow rate among different samples and its relationship with other factors.

[0037] Step S542, using the normalized data of the coolant pressure to construct the second coordinate axis. Specifically, in the data preparation link, it is necessary to extract the coolant pressure recorded time series data contained in each sample from the existing cooling sample set. The data can reflect the changes in the coolant pressure during the corresponding cutting process. Subsequently, perform normalization operations on the coolant pressure data. Similar to the normalization of the coolant flow rate data, according to the known minimum and maximum values of the original pressure data, determine the normalized result of a certain pressure value through the corresponding formula, so that all coolant pressure data are converted into the range of [0, 1]. In the process of generating the coordinate axis, in the same coordinate space selected before, with the vertical direction as the reference, set the value corresponding to the normalized coolant pressure data as the ordinate value to draw the coordinate axis, which is the second coordinate axis. Like the first coordinate axis, reasonably divide the scale interval according to the data characteristics and analysis accuracy requirements (for example, each 0.1 is a scale unit), so that the value of each cooling sample in the dimension of the coolant pressure can accurately find the corresponding position on this coordinate axis. After the construction of the second coordinate axis is completed, it is perpendicular to the first coordinate axis and jointly provides the necessary dimensional support for the configuration of the two-dimensional virtual coordinate system, which is conducive to subsequent comprehensive analysis of the mutual relationship between the coolant flow rate, pressure and cutting temperature and other factors from a two-dimensional perspective.

[0038] Step S543: Configure the two-dimensional virtual coordinate system according to the first coordinate axis and the second coordinate axis. Specifically, after separately constructing the first coordinate axis representing the normalized coolant flow rate data and the second coordinate axis representing the normalized coolant pressure data, integrate the coordinate systems, integrate them into the same coordinate plane, make them perpendicular to each other and intersect at the coordinate origin (0, 0), where this origin represents the normalized minimum state of the coolant flow rate and pressure. In this way, the two-dimensional virtual coordinate system is configured. In the coordinate system, any point in the plane can be represented by an ordered pair, and the elements in the pair respectively correspond to the normalized coolant flow rate and pressure data. Each cooling sample can find a unique coordinate position according to its corresponding normalized value, thereby presenting the complex cooling sample data in an intuitive two-dimensional plane form, providing a clear framework and basis for analysis operations such as center point sorting, and facilitating the exploration of the optimal combination of flow rate and pressure to control the cutting temperature. To make the two-dimensional virtual coordinate system clearer and easier to use, annotation and improvement work also needs to be carried out. Mark the corresponding variable names on the coordinate axes. For example, the first coordinate axis is marked as "Coolant Flow Rate (Normalized)", and the second coordinate axis is marked as "Coolant Pressure (Normalized)", and note the original data unit. At the same time, add a title such as "Two-Dimensional Distribution Coordinate System of Coolant Flow Rate - Pressure of Cooling Samples" to clarify the purpose and the data relationship shown. In addition, add auxiliary description information or legends as needed to facilitate subsequent operations such as observing sample distribution, data analysis, and comparing and communicating with other results, ensuring that it can effectively serve the cooling control optimization goal and helping to obtain the recommended coolant flow rate time series information and the recommended coolant pressure time series information.

[0039] In a possible implementation manner, perform center point sorting according to the two-dimensional virtual coordinate system to obtain the recommended coolant flow rate time series information and the recommended coolant pressure time series information. Step S550 further includes step S551: obtain the cooling sample distribution coordinate set of the cooling sample set. Specifically, based on the previously constructed two-dimensional virtual coordinate system, its abscissa is the normalized coolant flow rate data, and its ordinate is the normalized coolant pressure data. In the coordinate system, each cooling sample corresponds to a specific coordinate point, which is determined by the normalized values of its coolant flow rate and pressure. From the cooling sample set, extract the coordinate values of each sample in the two-dimensional virtual coordinate system one by one, and organize the coordinate values into a data set, that is, the cooling sample distribution coordinate set. Each element in the set is a two-dimensional vector in the form of , where represents the normalized value of the coolant flow rate, Represents the normalized value of the coolant pressure. In the above way, the distribution of the cooling samples in the two-dimensional space is presented in the form of a coordinate set, providing a data basis for subsequent outlier analysis, so as to further explore the distribution laws and characteristics of the cooling samples, identify possible abnormal samples, thereby optimizing the selection of coolant parameters, achieving precise control of the cutting temperature, improving the quality and efficiency of wafer cutting, reducing problems such as chip defects and material waste caused by temperature runaway, and ensuring the stability and reliability of the semiconductor manufacturing process.

[0040] Step S552, perform LOF outlier analysis on the cooling sample distribution coordinate set to obtain a cooling sample outlier factor set. Specifically, use the LOF (Local_Outlier_Factor) algorithm to perform outlier analysis on the cooling sample distribution coordinate set. This algorithm determines whether a sample point is an outlier based on the local density around the sample point. For each sample point in the coordinate set , calculate its distance to other sample points, usually using the Euclidean distance formula , where and are the coordinates of two sample points. Then, based on the distance information, determine the Local_Reachability_Density (LRD) of each sample point.

[0041] LRD reflects the local density situation around the sample point. The calculation method is relatively complex and involves concepts such as the k-distance neighborhood of the sample point. Briefly speaking, within a certain neighborhood of the sample point, considering the distribution of other sample points, calculate its reachability density. Finally, calculate the LOF value, that is, the outlier factor, for each sample point according to the LRD value. The larger the outlier factor, the greater the distribution difference between the sample point and other sample points in the local area, and the more likely it is to be an outlier. By performing such calculations on the entire cooling sample distribution coordinate set, a cooling sample outlier factor set is obtained, where each element corresponds to the outlier factor of the corresponding sample in the original cooling sample set. These outlier factors will serve as important bases for judging whether a sample is abnormal, helping us screen out samples that do not conform to the overall distribution law, making subsequent analysis and coolant parameter recommendations more accurate and reliable, avoiding interference from abnormal samples, improving the accuracy and stability of cooling control optimization, and providing a more scientific and reasonable coolant flow rate and pressure recommendation scheme for temperature control in the wafer cutting process.

[0042] Step S553: Extract the concentrated cooling samples in the cooling sample outlier factor set that are less than or equal to the outlier factor threshold. Specifically, after obtaining the cooling sample outlier factor set, it is necessary to determine a suitable outlier factor threshold. The threshold can be determined by various methods. For example, based on statistical methods, observing the distribution of outlier factors, combining practical experience and understanding of the data, select a value that can reasonably distinguish normal samples from abnormal samples. It can also be determined through experiments and simulations by analyzing the sample screening results under different thresholds and evaluating their impact on the coolant parameter recommendations, so as to determine an optimal threshold. Once the outlier factor threshold is determined, screen out the cooling samples corresponding to the elements that are less than or equal to this threshold from the cooling sample outlier factor set. These samples constitute the concentrated cooling sample set. The distribution of these samples in the two-dimensional virtual coordinate system is relatively concentrated, representing typical situations under normal cooling conditions. Their coolant flow rate and pressure combinations have high reference value and can reflect the range of coolant parameters that can effectively control the cutting temperature in most cases. By extracting these concentrated cooling samples, outlier samples caused by possible measurement errors, special working conditions or other abnormal factors are excluded, making the subsequent mean analysis more accurately reflect the optimal parameter trend under normal cooling conditions, providing a strong guarantee for obtaining reliable coolant recommended flow rate time series information and coolant recommended pressure time series information, helping to improve the temperature stability during wafer cutting, reducing the impact of temperature fluctuations on chip quality, and enhancing the overall process level of semiconductor manufacturing.

[0043] Step S554: Conduct a mean analysis on the concentrated cooling samples to obtain the coolant recommended flow rate time series information and the coolant recommended pressure time series information. Specifically, for the extracted concentrated cooling samples, calculate the mean values of their coolant flow rate recorded time series data and coolant pressure recorded time series data respectively. When calculating the mean value, since each cooling sample contains coolant flow rate and pressure data at different time nodes during the cutting process, it is necessary to perform point-by-point averaging on these time series data. For example, for the samples in the concentrated cooling sample set, the coolant flow rate time series data of the th sample is (where represents time), then the mean value of the coolant flow rate at a certain moment can be calculated by the formula Calculated. Similarly, for the coolant pressure time series data, a similar method is used to calculate the mean value. After such mean value calculation, a set of coolant recommended flow rate time series information and a set of coolant recommended pressure time series information are obtained. These recommended information reflect the optimal change trends of the coolant flow rate and pressure as the cutting process progresses under normal cooling conditions, and can provide precise control parameters for the actual wafer cutting cooling control system, enabling it to dynamically adjust the coolant flow rate and pressure according to the real-time situation of the cutting process, ensuring that the cutting temperature is always maintained within an appropriate range, improving the quality and stability of wafer cutting, reducing chip defects and material damage caused by temperature problems, providing a high-quality wafer basis for subsequent semiconductor manufacturing processes, enhancing the production efficiency and economic benefits of the entire semiconductor manufacturing process, and at the same time providing data support and theoretical basis for the optimization and improvement of cooling control strategies, contributing to further improving the energy utilization efficiency and process control level in the semiconductor manufacturing process.

[0044] In a possible implementation manner, when the number of samples in the cooling sample set is greater than the cooling sample number threshold, according to the coolant flow rate recorded time series data and the coolant pressure recorded time series data, the cooling sample set is distributed in a two-dimensional virtual coordinate system. Step S540 further includes step S544. When the number of samples in the cooling sample set is less than or equal to the cooling sample number threshold, return to execute the loop of updating the cutting speed time series information and the cutting depth time series information. Specifically, clarify the number of samples in the current cooling sample set and compare it with the pre-set cooling sample number threshold. This threshold is determined by comprehensively considering factors such as the complexity of the subsequent data analysis model, the optimization accuracy of coolant parameters, past experience, and the understanding of the relationship between cutting variables. If the number of samples is less than or equal to the threshold, it indicates that the data is insufficient to support subsequent operations, and it is necessary to return to update the cutting speed time series information and the cutting depth time series information. For the cutting speed, based on experience and theoretical analysis, appropriately expand the value range, such as from [500, 1000] millimeters per minute to [300, 1200] millimeters, and refine the value interval from 100 millimeters per minute to 50 millimeters per minute; for the cutting depth, similarly, such as from [0.1, 0.5] millimeters to [0.05, 0.6] millimeters, and the interval is refined from 0.1 millimeter to 0.05 millimeter to enrich the data change situation. Then, execute the loop again, re-conduct the wafer cutting experiment according to the updated parameters, record data such as coolant and cutting temperature to form new samples and add them to the set, and then continue to compare the number of samples with the threshold, and repeat the loop until the number of samples exceeds the threshold, so as to provide a sufficient and reliable data basis for obtaining the coolant recommended flow rate time series information and the coolant recommended pressure time series information subsequently, realizing effective optimization of cooling control, and ensuring cutting quality and efficiency.

[0045] Step S600: Perform wafer cutting according to the cutting speed timing information, the cutting depth timing information, the recommended coolant flow rate timing information, and the recommended coolant pressure timing information to obtain a number of separated chips for MOSFET co-linear integrated packaging. Specifically, analyze the cutting speed timing information, adjust the speed control system of the cutting equipment to make it move according to a predetermined speed curve; study the cutting depth timing information, accurately set the depth control parameters of the cutting equipment to ensure that the cutting depth accurately follows the predetermined change. Input the recommended coolant flow rate and pressure timing information into the cooling control system, debug and calibrate the coolant supply system to make it stably and accurately spray coolant into the cutting area. Then start the wafer cutting equipment, the cutting tool cuts the wafer according to the preset speed and depth timing, and the coolant cools synchronously. After completion, carefully separate the chips and strictly detect the quality, including dimensional accuracy, surface quality, and electrical performance, etc. The qualified chips are processed according to the MOSFET co-linear integrated packaging process flow, through operations such as chip mounting, wire bonding, potting, etc., strictly following the process standards and quality control requirements to ensure that high-quality and high-performance products are finally produced to meet the market demand and promote the development of the industry.

[0046] In the above text, with reference to Figure 1 a MOSFET co-linear packaging control method according to an embodiment of the present invention was described in detail. Next, with reference to Figure 2 a MOSFET co-linear packaging control system according to an embodiment of the present invention will be described.

[0047] A MOSFET co-linear packaging control system according to an embodiment of the present invention solves the technical problems of inaccurate temperature control and low cooling efficiency in the wafer cutting process in the prior art, and achieves the technical effects of improving the accuracy of temperature control and the cooling efficiency in the wafer cutting process. A MOSFET co-linear packaging control system includes: a chip distribution frame diagram acquisition module 10, a cutting shortest path optimization module 20, a timing information acquisition module 30, a cutting temperature timing information acquisition module 40, a cooling control optimization module 50, and a wafer cutting module 60.

[0048] The chip distribution frame diagram acquisition module 10 is used to obtain the chip distribution frame diagram of the wafer to be cut.

[0049] The cutting shortest path optimization module 20 is used to optimize the cutting shortest path according to the chip distribution frame diagram to obtain a recommended wafer cutting path.

[0050] The timing information acquisition module 30 is used to obtain the cutting speed timing information and the cutting depth timing information.

[0051] The cutting temperature time series information acquisition module 40 is used to process the wafer cutting recommended path, the cutting speed time series information, and the cutting depth time series information through a cutting temperature prediction model to obtain cutting temperature time series information.

[0052] The cooling control optimization module 50 is used to optimize the cooling control according to the cutting temperature time series information to obtain the recommended coolant flow rate time series information and the recommended coolant pressure time series information.

[0053] The wafer cutting module 60 is used to perform wafer cutting according to the cutting speed time series information, the cutting depth time series information, the recommended coolant flow rate time series information, and the recommended coolant pressure time series information to obtain a plurality of separated chips for MOSFET co-line integrated packaging.

[0054] Next, the specific configuration of the cutting temperature time series information acquisition module 40 will be described in detail. As described above, through the cutting temperature prediction model, the wafer cutting recommended path, the cutting speed time series information, and the cutting depth time series information are processed to obtain the cutting temperature time series information. The cutting temperature time series information acquisition module 40 further includes: a wafer cutting equipment parameter acquisition unit, which is used to obtain wafer cutting equipment parameters, where the wafer cutting equipment parameters include cutting channel structure parameters, cutting groove structure parameters, connection rib structure parameters, and cutting machine accuracy parameters; a cutting temperature prediction model construction unit, which is used to construct the cutting temperature prediction model with the cutting channel structure parameters, the cutting groove structure parameters, the connection rib structure parameters, and the cutting machine accuracy parameters as constraints; a cutting temperature time series information acquisition unit, which is used to process the wafer cutting recommended path, the cutting speed time series information, and the cutting depth time series information through the cutting temperature prediction model to obtain the cutting temperature time series information.

[0055] Among them, a cutting temperature prediction model is constructed with the cutting channel structure parameters, the cutting groove structure parameters, the connecting rib structure parameters, and the cutting machine precision parameters as constraints. The cutting temperature prediction model construction unit further includes: a similarity threshold configuration subunit for configuring a cutting channel structure similarity threshold, a cutting groove structure similarity threshold, a connecting rib structure similarity threshold, and a cutting machine precision deviation distance threshold; a first cutting sample acquisition subunit for obtaining a first cutting sample, where the first cutting sample has a cutting channel structure label, a cutting groove structure label, a connecting rib structure label, and a cutting machine precision label; a similarity acquisition subunit for comparing the cutting channel structure label, the cutting groove structure label, the connecting rib structure label, and the cutting machine precision label with the cutting channel structure parameters, the cutting groove structure parameters, the connecting rib structure parameters, and the cutting machine precision parameters to obtain a cutting channel structure similarity, a cutting groove structure similarity, a connecting rib structure similarity, and a cutting machine precision deviation distance; a sample set construction subunit for adding the first cutting sample into the cutting temperature prediction model construction sample set when the cutting channel structure similarity is greater than or equal to the cutting channel structure similarity threshold, the cutting groove structure similarity is greater than or equal to the cutting groove structure similarity threshold, the connecting rib structure similarity is greater than or equal to the connecting rib structure similarity threshold, and the cutting machine precision deviation distance is less than or equal to the cutting machine precision deviation distance threshold; a cutting temperature prediction model training subunit for training the cutting temperature prediction model when the cutting temperature prediction model construction sample set is greater than or equal to a preset sample quantity.

[0056] Next, the specific configuration of the cooling control optimization module 50 will be described in detail. As described above, the cooling control is optimized according to the cutting temperature timing information to obtain the coolant recommended flow rate timing information and the coolant recommended pressure timing information. The cooling control optimization module 50 further includes: an ambient temperature monitoring value acquisition unit for acquiring an ambient temperature monitoring value; a wafer cutting equipment parameter composition unit for acquiring wafer cutting equipment parameters, where the wafer cutting equipment parameters include cutting channel structure parameters, cutting groove structure parameters, connection rib structure parameters, and cutting machine accuracy parameters; a cooling sample set acquisition unit for collecting cooling samples with the cutting channel structure parameters, the cutting groove structure parameters, the connection rib structure parameters, the cutting machine accuracy parameters, and the cutting temperature timing information as constraints to obtain a cooling sample set, where any one cooling sample has coolant flow rate recording timing data and coolant pressure recording timing data; a coordinate system distribution unit for distributing the cooling sample set in a two-dimensional virtual coordinate system according to the coolant flow rate recording timing data and the coolant pressure recording timing data when the number of samples in the cooling sample set is greater than a cooling sample number threshold; and a center point sorting unit for sorting center points according to the two-dimensional virtual coordinate system to obtain the coolant recommended flow rate timing information and the coolant recommended pressure timing information.

[0057] Among them, when the number of samples in the cooling sample set is greater than a cooling sample number threshold, the cooling sample set is distributed in a two-dimensional virtual coordinate system according to the coolant flow rate recording timing data and the coolant pressure recording timing data. The coordinate system distribution unit further includes: a first coordinate axis construction subunit for constructing a first coordinate axis with the normalized data of the coolant flow rate; a second coordinate axis construction subunit for constructing a second coordinate axis with the normalized data of the coolant pressure; and a virtual coordinate system configuration subunit for configuring the two-dimensional virtual coordinate system according to the first coordinate axis and the second coordinate axis.

[0058] Among them, according to the two-dimensional virtual coordinate system for center point sorting, the coolant recommended flow rate time series information and the coolant recommended pressure time series information are obtained. The center point sorting unit further includes: a distribution coordinate set acquisition subunit, which is used to obtain the cooling sample distribution coordinate set of the cooling sample set; an outlier analysis subunit, which is used to perform LOF outlier analysis on the cooling sample distribution coordinate set to obtain a cooling sample outlier factor set; a concentrated cooling sample extraction subunit, which is used to extract the concentrated cooling samples in the cooling sample outlier factor set that are less than or equal to the outlier factor threshold; and an average value analysis subunit, which is used to perform average value analysis on the concentrated cooling samples to obtain the coolant recommended flow rate time series information and the coolant recommended pressure time series information.

[0059] Among them, when the number of samples in the cooling sample set is greater than the cooling sample number threshold, according to the coolant flow rate record time series data and the coolant pressure record time series data, the cooling sample set is distributed in a two-dimensional virtual coordinate system. The coordinate system distribution unit further includes: an execution loop subunit, which is used to return and update the cutting speed time series information and the cutting depth time series information execution loop when the number of samples in the cooling sample set is less than or equal to the cooling sample number threshold.

[0060] A MOSFET collinear packaging control system provided by an embodiment of the present invention can execute a MOSFET collinear packaging control method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

[0061] Although this application makes various references to certain modules in the system according to the embodiments of this application, however, any number of different modules can be used and run on the user terminal and / or the server. The included various units and modules are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0062] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be executed in an order different from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A MOSFET co-linear packaging control method, characterized in that: include: Obtain a chip distribution frame diagram of the wafer to be cut; Optimize the shortest cutting path according to the chip distribution frame diagram to obtain a recommended wafer cutting path; Obtaining cutting speed timing information and cutting depth timing information; The wafer cutting recommended path, the cutting speed timing information and the cutting depth timing information are processed by a cutting temperature prediction model to obtain cutting temperature timing information; Optimizing the cooling control according to the cutting temperature timing information to obtain the recommended coolant flow timing information and the recommended coolant pressure timing information; Wafer cutting is performed according to the cutting speed timing information, the cutting depth timing information, the coolant recommended flow timing information and the coolant recommended pressure timing information to obtain a plurality of separate chips for MOSFET co-linear integration packaging.

2. The method according to claim 1, characterized in that The wafer cutting recommended path, the cutting speed information and the cutting depth timing information are processed by a cutting temperature prediction model to obtain cutting temperature timing information, including: Obtaining wafer cutting equipment parameters, wherein the wafer cutting equipment parameters include cutting road structure parameters, cutting groove structure parameters, connecting rib structure parameters, and cutting machine precision parameters; The cutting temperature prediction model is constructed with the cutting path structural parameters, the cutting groove structural parameters, the connecting rib structural parameters, and the cutting machine precision parameters as constraints; The wafer cutting recommended path, the cutting speed timing information and the cutting depth timing information are processed by a cutting temperature prediction model to obtain cutting temperature timing information.

3. The method according to claim 2, characterized in that The cutting temperature prediction model is constructed with the cutting path structural parameters, the cutting groove structural parameters, the connecting rib structural parameters, and the cutting machine precision parameters as constraints, including: Configure the cutting path structure similarity threshold, cutting groove structure similarity threshold, connecting rib structure similarity threshold and cutting machine accuracy deviation distance threshold; Obtaining a first cutting sample, wherein the first cutting sample has a cutting track structure label, a cutting groove structure label, a connecting rib structure label, and a cutting machine precision label; Compare the cutting path structure label, the cutting groove structure label, the connecting rib structure label, and the cutting machine precision label with the cutting path structure parameters, the cutting groove structure parameters, the connecting rib structure parameters, and the cutting machine precision parameters to obtain the cutting path structure similarity, the cutting groove structure similarity, the connecting rib structure similarity, and the cutting machine precision deviation distance; When the cutting road structure similarity is greater than or equal to the cutting road structure similarity threshold, the cutting groove structure similarity is greater than or equal to the cutting groove structure similarity threshold, the connecting rib structure similarity is greater than or equal to the connecting rib structure similarity threshold, and the cutting machine precision deviation distance is less than or equal to the cutting machine precision deviation distance threshold, the first cutting sample is added into the cutting temperature prediction model to construct the sample set; When the cutting temperature prediction model construction sample set is greater than or equal to the preset sample number, the cutting temperature prediction model is trained.

4. The method according to claim 1, characterized in that Optimizing the cooling control according to the cutting temperature timing information to obtain the coolant recommended flow timing information and the coolant recommended pressure timing information, including: Get the ambient temperature monitoring value; Obtaining wafer cutting equipment parameters, wherein the wafer cutting equipment parameters include cutting road structure parameters, cutting groove structure parameters, connecting rib structure parameters, and cutting machine precision parameters; Cooling samples are collected with the cutting path structural parameters, the cutting groove structural parameters, the connecting rib structural parameters, the cutting machine precision parameters, and the cutting temperature timing information as constraints to obtain a cooling sample set, wherein any cooling sample has coolant flow recording timing data and coolant pressure recording timing data; When the number of samples of the cooling sample set is greater than a cooling sample number threshold, distributing the cooling sample set in a two-dimensional virtual coordinate system according to the coolant flow recording time series data and the coolant pressure recording time series data; The center points are sorted according to the two-dimensional virtual coordinate system to obtain the coolant recommended flow time series information and the coolant recommended pressure time series information.

5. The method according to claim 4, characterized in that According to the coolant flow recording time series data and the coolant pressure recording time series data, the cooling sample set is distributed in a two-dimensional virtual coordinate system, including: The first coordinate axis is constructed using the normalized data of the coolant flow rate; The second coordinate axis is constructed using the normalized data of the cooling fluid pressure; The two-dimensional virtual coordinate system is configured according to the first coordinate axis and the second coordinate axis.

6. The method according to claim 4, characterized in that The center point is sorted according to the two-dimensional virtual coordinate system to obtain the coolant recommended flow time series information and the coolant recommended pressure time series information, including: Obtaining a cooling sample distribution coordinate set of the cooling sample set; Performing LOF outlier analysis on the cooling sample distribution coordinate set to obtain a cooling sample outlier factor set; Extracting concentrated cooling samples whose outlier factors are less than or equal to the outlier factor threshold value from the cooling sample outlier factor set; The centralized cooling samples are subjected to mean analysis to obtain the coolant recommended flow time series information and the coolant recommended pressure time series information.

7. The method according to claim 4, characterized in that Also includes: When the sample quantity of the cooling sample set is less than or equal to the cooling sample quantity threshold, the loop is executed by returning to update the cutting speed timing information and the cutting depth timing information.

8. A MOSFET co-line packaging control system, characterized in that: The system is used to implement a MOSFET co-linear packaging control method according to any one of claims 1 to 7, and the system comprises: A chip distribution frame diagram acquisition module, wherein the chip distribution frame diagram acquisition module is used to obtain a chip distribution frame diagram of a wafer to be cut; A cutting shortest path optimization module, wherein the cutting shortest path optimization module is used to optimize the cutting shortest path according to the chip distribution frame line diagram to obtain a wafer cutting recommended path; A timing information acquisition module, wherein the timing information acquisition module is used to obtain cutting speed timing information and cutting depth timing information; A cutting temperature timing information acquisition module, wherein the cutting temperature timing information acquisition module is used to process the wafer cutting recommended path, the cutting speed timing information and the cutting depth timing information through a cutting temperature prediction model to obtain cutting temperature timing information; A cooling control optimization module, the cooling control optimization module is used to optimize the cooling control according to the cutting temperature timing information, and obtain the coolant recommended flow timing information and the coolant recommended pressure timing information; A wafer cutting module is used to cut wafers according to the cutting speed timing information, the cutting depth timing information, the coolant recommended flow timing information and the coolant recommended pressure timing information to obtain a plurality of separate chips for MOSFET co-linear integration packaging.

Images