Process dynamic control method and device for silicon carbide high-voltage MOSFET

Abstract

The invention provides a process dynamic control method and device for a silicon carbide high-voltage MOSFET, and relates to the technical field of semiconductor manufacturing, and the method comprises the steps: carrying out the production process analysis of the silicon carbide high-voltage MOSFET, and building a production process stage; calling a historical production data set by using a production process stage to carry out time sequence correlation analysis, and creating an exception propagation network; establishing an intelligent management platform based on the anomaly propagation network; after real-time production data of the edge sensor is read by using the intelligent management platform, abnormal propagation analysis is carried out based on an abnormal propagation network, and production feedback attention is established; and process dynamic optimization management of the production process stage is carried out through production feedback attention. The technical problem that an existing process control method neglects mutual influence between different production stages, so that abnormity in the production process is difficult to find and control in time is solved, and the production quality and the production efficiency are improved.

Description

Technical Field

[0001] The present application relates to the field of semiconductor manufacturing technology, and in particular to a process dynamic control method and device for silicon carbide high-voltage MOSFET. Background Art

[0002] Silicon carbide (SiC) high-voltage MOSFETs are widely used in power conversion equipment in harsh working environments such as high voltage, high frequency, and high temperature due to their advantages such as high withstand voltage and low on-resistance. The production process of silicon carbide high-voltage MOSFETs is complex and involves multiple process stages, such as pre-treatment, epitaxial growth, ion implantation and diffusion, metallization and contact structure formation, and packaging testing, each of which has an important impact on the final device performance. Existing silicon carbide high-voltage MOSFET process control methods mainly focus on each independent production process stage. In these methods, the process control of each stage is relatively independent, lacking in-depth consideration of the correlation between the various stages of the entire production process, and often ignoring the mutual influence between different process stages. During the production process, anomalies or changes in the previous stage will accumulate to the next stage, causing anomalies or failures in the next stage, resulting in low production efficiency and unstable product quality. Summary of the invention

[0003] The present application provides a process dynamic control method and device for silicon carbide high-voltage MOSFET, which solves the technical problem that the existing process control method is limited to the optimization of a single stage, ignores the mutual influence between different production stages, and makes it difficult to timely detect and control abnormalities in the production process, thereby affecting production efficiency and product quality, and achieves the technical effect of improving the production quality and production efficiency of silicon carbide high-voltage MOSFET.

[0004] In view of the above problems, on the one hand, the present application provides a process dynamic control method for silicon carbide high-voltage MOSFET, the method comprising: performing production process analysis on silicon carbide high-voltage MOSFET, establishing production process stages, the production process stages comprising a pre-processing stage, an epitaxial growth stage, an ion implantation and diffusion stage, a metallization and contact structure formation stage, and a packaging and testing stage; using the production process stages to call historical production data sets, performing timing correlation analysis based on the historical production data sets, and creating an abnormal propagation network using the timing correlation analysis results; establishing an intelligent management platform based on the abnormal propagation network, the intelligent management platform digitally communicating with an edge sensor group distributed in each production process stage, the edge sensor being used to collect production data at each production process stage; after using the intelligent management platform to read the real-time production data of the edge sensor, performing abnormal propagation analysis based on the abnormal propagation network, and establishing a production feedback focus; and performing process dynamic optimization management of the production process stage through the production feedback focus.

[0005] On the other hand, the present application also provides a process dynamic control device for silicon carbide high-voltage MOSFET, the device comprising: a production process analysis module, used to perform production process analysis on silicon carbide high-voltage MOSFET and establish production process stages, the production process stages comprising a pre-processing stage, an epitaxial growth stage, an ion implantation and diffusion stage, a metallization and contact structure formation stage, and a packaging and testing stage; a historical production data analysis module, used to call historical production data sets using the production process stages, perform timing correlation analysis based on the historical production data sets, and create an abnormal propagation network using the timing correlation analysis results; a management platform construction module, used to establish an intelligent management platform based on the abnormal propagation network, the intelligent management platform digitally communicates with an edge sensor group distributed in each production process stage, the edge sensor is used to collect production data of each production process stage; an abnormal propagation analysis module, used to perform abnormal propagation analysis based on the abnormal propagation network after reading the real-time production data of the edge sensor using the intelligent management platform, and establish a production feedback focus; a process dynamic optimization module, used to perform process dynamic optimization management of the production process stage through the production feedback focus.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: By analyzing the production process of silicon carbide high-voltage MOSFET in detail, it is divided into multiple process stages, so that each process stage can be managed independently and provide a clear framework for the optimization of the overall process. Using historical production data sets for time series correlation analysis can reveal the potential correlation between the process stages and discover potential problems that may exist in the process. Time series analysis can not only review historical data, but also predict possible abnormal propagation paths, identify potential fault hazards in advance, and avoid the spread of problems. Based on the results of time series correlation analysis, an abnormal propagation network is established. The abnormal propagation network can intuitively present the propagation path of abnormal situations between stages, so that abnormalities in the production process can be effectively tracked and located, the source of the problem can be accurately identified, and data support can be provided for subsequent intervention and adjustment. Through the intelligent management platform, digital communication is carried out with edge sensors distributed in various production process stages to collect real-time data from each process stage, ensuring that each link in the production process can obtain real-time feedback, providing a basis for subsequent data analysis and decision-making. In the intelligent management platform, after reading the real-time production data of the edge sensor, the abnormal propagation analysis is combined with the abnormal propagation network to finally form production feedback attention. Through the analysis of real-time data, we can further focus on abnormal situations in the production process, accurately judge the impact range and propagation direction of abnormalities based on the abnormal propagation network, and provide accurate basis for process optimization. Through production feedback, we can further carry out dynamic optimization management of the production process stage to ensure that the production process is always in the best state.

[0007] To summarize, this application creates an abnormal propagation network through a comprehensive analysis of the silicon carbide high-voltage MOSFET production process, establishes an intelligent management platform to communicate with the edge sensor group, collects production data in real time for abnormal propagation analysis, and realizes dynamic optimization of the entire production process and real-time abnormality warning, thereby improving the production efficiency and product quality of silicon carbide high-voltage MOSFET, and enhancing the product's performance and reliability in harsh working environments such as high voltage, high frequency, and high temperature.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A schematic flow chart of a process dynamic control method for a silicon carbide high-voltage MOSFET provided in an embodiment of the present application.

[0010] Figure 2 A schematic diagram of a process flow for creating an abnormal propagation network in a process dynamic control method for silicon carbide high-voltage MOSFET provided in an embodiment of the present application.

[0011] Figure 3 A schematic diagram of the structure of a process dynamic control device for silicon carbide high-voltage MOSFET provided in an embodiment of the present application.

[0012] Explanation of the reference numerals: production process analysis module 10 , historical production data analysis module 20 , management platform construction module 30 , abnormal propagation analysis module 40 , process dynamic optimization module 50 . DETAILED DESCRIPTION

[0013] The embodiment of the present application provides a process dynamic control method and device for silicon carbide high-voltage MOSFET, divides the silicon carbide high-voltage MOSFET production process into production process stages, and calls historical production data sets for time series correlation analysis to create an abnormal propagation network. Real-time production data is acquired, and abnormal propagation analysis is performed on the real-time production data based on the abnormal propagation network. A production feedback focus is established, and dynamic process optimization management of the production process stage is performed through the production feedback focus. The embodiment of the present application solves the technical problem that the existing process control method is limited to the optimization of a single stage, ignoring the mutual influence between different production stages, resulting in difficulty in timely detection and control of abnormalities in the production process, thereby affecting production efficiency and product quality, and achieves the technical effect of improving the production quality and production efficiency of silicon carbide high-voltage MOSFET.

[0014] Embodiment 1, as Figure 1 As shown, an embodiment of the present application provides a process dynamic control method for silicon carbide high-voltage MOSFET, the method comprising: Step S1: Analyze the production process of silicon carbide high-voltage MOSFET and establish production process stages, which include pre-processing stage, epitaxial growth stage, ion implantation and diffusion stage, metallization and contact structure formation stage, and packaging and testing stage.

[0015] Specifically, the production process stage refers to each specific stage in the production process according to the process flow. The production process of silicon carbide high-voltage MOSFET is analyzed in detail, and the entire production process is divided into several key process stages, including the pre-treatment stage, epitaxial growth stage, ion implantation and diffusion stage, metallization and contact structure formation stage, packaging and testing stage, etc. The process conditions, material usage, temperature, pressure and other factors in each stage have an important impact on the quality of the final product. Among them, the pre-treatment stage includes operations such as cleaning and surface treatment of the silicon carbide substrate material to provide a clean and suitable surface for subsequent epitaxial growth and other steps. In the epitaxial growth stage, an epitaxial layer is grown on the silicon carbide substrate by chemical vapor deposition (CVD) technology. The doping concentration and thickness of the epitaxial layer have an important influence on the electrical properties of the MOSFET. The ion implantation and diffusion stage is to implant specific ions (such as boron, phosphorus, etc.) into the silicon carbide substrate, and then diffuse the ions evenly in the silicon crystal by heating to form the required semiconductor structure. The metallization and contact structure formation stage is to form an electrical connection by depositing a metal layer to ensure that the current can flow through the MOSFET. The packaging and testing stage is to encapsulate the manufactured silicon carbide high-voltage MOSFET chip in a protective shell to provide mechanical protection, heat dissipation and other functions, and at the same time lead out the electrodes for performance testing and final quality inspection.

[0016] By analyzing the production process, a structured perspective is provided for the entire process, and the complex production process is managed in a modular manner, which facilitates subsequent production process optimization, quality control and problem troubleshooting.

[0017] Step S2: using the production process stage to call a historical production data set, performing a time series correlation analysis based on the historical production data set, and creating an abnormal propagation network using the time series correlation analysis result.

[0018] Specifically, historical production data sets refer to relevant data collected during past production processes, including operating parameters, environmental conditions, production time, product performance, etc. at each process stage. Time series correlation analysis is an analysis method based on time series data. By analyzing the correlation between data at different time points, factors and potential trends that affect the process are identified. Anomaly propagation network is a network model built based on the results of time series correlation analysis, which is used to represent the propagation path and relationship of abnormal situations between various stages in the production process.

[0019] Using the production process stage established in step S1 as an index, call the historical production data set from the production database. Use data mining and analysis methods, such as association rule algorithms, time series analysis models, etc., to analyze the data of different process stages in the historical production data set on the time axis, and mine the temporal correlation between the data. For example, analyze the impact of the change of a certain parameter in the pre-processing stage on the product quality of the epitaxial growth stage after a period of time. Then, based on the results of these temporal correlation analyses, determine the mutual influence relationship between the various production process stages, and use network modeling tools (such as Graphviz, etc.) to create an abnormal propagation network. This network uses the production process stage as a node and the abnormal propagation relationship between the various production process stages as an edge.

[0020] By constructing an anomaly propagation network through time-series correlation analysis, the anomaly propagation path can be clarified, so as to accurately locate the source of the anomaly and optimize the process in subsequent production.

[0021] Step S3: An intelligent management platform is established based on the abnormal propagation network. The intelligent management platform digitally communicates with edge sensor groups distributed in each production process stage. The edge sensors are used to collect production data at each production process stage.

[0022] Specifically, the intelligent management platform is a system platform that integrates data collection, processing, analysis and control. It can monitor and manage the production process in real time and make process adjustments and optimizations. Edge sensors are sensors distributed in various production process stages during the production process, responsible for real-time collection of physical, chemical or mechanical parameters (such as temperature, humidity, voltage, etc.). Multiple edge sensors in each production process stage form an edge sensor group. The edge sensor group and the intelligent management platform exchange data through digital signals to ensure the real-time and accuracy of data transmission.

[0023] Based on the abnormal propagation network established in the previous step, an intelligent management platform is built. The platform is able to receive edge sensor data from each production process stage in real time. These edge sensor groups can monitor different parameters in each production process stage, such as temperature sensors monitoring temperature changes in the epitaxial growth stage, pressure sensors monitoring pressure changes in the metallization stage, and current sensors monitoring current fluctuations in the ion implantation stage. These data are transmitted to the intelligent management platform in real time through digital communication for centralized processing and analysis, ensuring that the platform can instantly understand the status of each production stage and respond to abnormal situations.

[0024] Through the intelligent management platform and edge sensor group, full digital monitoring of the production process is achieved, so that data from each link can be transmitted to the intelligent platform in real time, providing timely data support for anomaly detection and production optimization.

[0025] Step S4: After using the intelligent management platform to read the real-time production data of the edge sensor, an abnormal propagation analysis is performed based on the abnormal propagation network to establish production feedback attention.

[0026] Specifically, real-time production data is data collected in real time by edge sensors, reflecting the real-time status of each link in the production process. Production feedback focuses on the links or parameters that need attention and optimization in the production process based on the results of abnormal propagation analysis, helping the production line to make adjustments.

[0027] After the intelligent management platform reads the real-time production data collected by the edge sensor, it inputs this data into the analysis model built based on the abnormal propagation network to determine the source, propagation path and impact range of the abnormal situation in the production process. This analysis model can be built based on a machine learning algorithm, such as a decision tree algorithm. Taking the ion implantation and diffusion stage as an example, if the real-time data shows that the dose of the implanted ions deviates from the set value, the abnormal propagation analysis model, combined with the abnormal propagation network, can analyze the possible impact of this deviation on the electrical contact performance in the subsequent metallization and contact structure formation stage, as well as the possible propagation path of this impact.

[0028] Through abnormal propagation analysis, the source of abnormalities in the production process can be quickly located, and the impact of abnormalities on the entire production process can be accurately evaluated, providing an accurate basis for subsequent process optimization.

[0029] Step S5: Dynamic optimization management of the process in the production process stage is performed through the production feedback attention.

[0030] Specifically, according to the production feedback results established in step S4, the parameters and operation procedures of each production process stage are dynamically adjusted and optimized. For example, if it is found that the temperature fluctuation in the epitaxial growth stage has a great impact on the product quality, then in the process dynamic optimization management, the temperature can be stabilized by adjusting the heating system parameters of the CVD equipment. If it is an injection dose problem in the ion implantation and diffusion stage, the working parameters of the ion implanter can be adjusted.

[0031] Through dynamic process optimization management, the production process can be adjusted in real time to ensure that the production process of silicon carbide high-voltage MOSFET is always in the best state, thereby improving the product yield and production efficiency.

[0032] Further, such as Figure 2 As shown, step S2 includes: Step S21: After mapping the production process stages of the historical production data set, data segmentation is performed on the historical production data set to establish active control data and key quality parameters, wherein the active control data includes temperature, pressure, time, and gas flow rate, and the key quality parameters include epitaxial growth rate and heat dissipation layer quality; Step S22: calling the autocorrelation channel and the cross-correlation channel to analyze the time lag relationship between the active control data and the key quality parameters, and establishing a timing lag association within the production process stage; Step S23: establishing a correlation analysis channel, using the correlation analysis channel to perform correlation analysis between production process stages, and establishing a time-series production correlation; Step S24: using the timing lag correlation and the timing production correlation to establish a timing correlation analysis result.

[0033] Specifically, the historical production data is mapped to each process stage so that each data item in the data set can correspond to a specific production stage. For example, data such as temperature and gas flow in the epitaxial growth stage need to be associated with the production conditions in the epitaxial growth process. Then, the historical data is divided into two types of data: active control data and key quality parameters. Among them, active control data are parameters that can be directly controlled and adjusted during the production process of silicon carbide high-voltage MOSFET, such as temperature, pressure, time, gas flow, etc. Key quality parameters are parameters that have a key impact on the quality of the final product of silicon carbide high-voltage MOSFET, such as epitaxial growth rate, heat dissipation layer quality, etc. By accurately mapping and segmenting historical production data, the foundation is laid for the subsequent analysis of the relationship between different data, which helps to more accurately identify the key factors and controllable production parameters that affect product quality.

[0034] The autocorrelation channel is a channel or method used to analyze the correlation within the same set of data at different time points. For example, for the temperature data in the active control data, the autocorrelation channel can analyze the correlation between the temperature data at different time points, reflecting the change pattern of the temperature data itself over time. The cross-correlation channel is a channel used to analyze the correlation between two different sets of data (active control data and key quality parameters) at different time points. For example, analyze the relationship between temperature (active control data) and epitaxial growth rate (key quality parameters) at different time points. The time-lag correlation is obtained based on the time-lag relationship analysis, reflecting the result of the time-lag correlation relationship between the active control data and the key quality parameters in the production process stage. For example, in a certain production process stage, it is found that after the pressure changes, it will take a certain period of time to affect the epitaxial growth rate. This correlation is the time-lag correlation.

[0035] Use statistical analysis software, such as MATLAB or the SciPy library in Python. For active control data and key quality parameters, perform time lag relationship analysis through autocorrelation channels and cross-correlation channels respectively. For example, for temperature data (active control data), the autocorrelation coefficient at different time lags is calculated through the autocorrelation function to determine the time lag relationship of the temperature data itself. For the two sets of data, temperature and epitaxial growth rate, the cross-correlation coefficient at different time lags is calculated through the cross-correlation function to determine the time lag relationship between them. Through these calculations and analyses, the timing lag association within the production process stage is established.

[0036] The association analysis channel is a tool or method used to analyze the association relationship between different production process stages. The association analysis channel is established using data mining algorithms, such as decision tree algorithms or association rule mining algorithms, and the data of different production process stages (including active control data and key quality parameters) are used as input and analyzed through the association analysis channel. For example, the data of the pre-processing stage and the data of the epitaxial growth stage are input into the association analysis channel to analyze the association relationship between these two types of data. Through this analysis, the time-series production association between different production process stages is established. By establishing the time-series production association, the dependency relationship between each process stage can be revealed, so that in the production process, it is easy to predict in advance how the adjustment of a certain stage will affect other stages, thereby providing a more comprehensive basis for process control and optimization.

[0037] The results of time series correlation analysis refer to the comprehensive analysis results obtained by combining autocorrelation analysis, cross-correlation analysis and correlation analysis between production process stages, which reveal the time relationship and interdependence of each process link in the production process. Integrating the analysis results of time series lag correlation and time series production correlation into the final time series correlation analysis results provides a comprehensive and accurate analysis framework that can dynamically monitor and predict the impact of each process stage in the production process, providing strong data support for abnormal prediction and process optimization.

[0038] Further, step S23 includes: Step S231: configure the correlation calculation function of the correlation analysis channel as follows: ;in, Characterization process stage and process stages The association, Represents the time point index, which is used to indicate the time sequence of data. is the total length of the time series, representing the total number of data points involved in the association calculation, is the time weighting factor, characterizing each time point The weight of Characterize active control data of the front-end process stage, Characterize key quality parameters of post-process stages, A nonlinear mapping function that characterizes the active control data of the preceding process stage, Nonlinear mapping functions that characterize key quality parameters of post-process stages, The square term of the nonlinear mapping function that characterizes the active control data, The square term of the nonlinear mapping function characterizing the key quality parameters.

[0039] Step S232: performing association analysis using the association calculation function to establish the time series production association.

[0040] Specifically, the correlation calculation function is configured for the correlation analysis channel as follows: ,in, Characterization process stage and process stages The association, Represents the time point index, which is used to indicate the time sequence of data. is the total length of the time series, representing the total number of data points involved in the association calculation, is the time weighting factor, characterizing each time point Since the production process is carried out over time, data at different time points may have different impacts on the final correlation results. The time weighting factor is used to adjust the degree of this impact. Characterize active control data of the front-end process stage, Characterize key quality parameters of post-process stages, A nonlinear mapping function that characterizes the active control data of the preceding process stage, Nonlinear mapping functions that characterize key quality parameters of post-process stages, The square term of the nonlinear mapping function that characterizes the active control data, The square term of the nonlinear mapping function characterizing the key quality parameter. For example, for the temperature in the epitaxial growth stage and the thickness of the ion implantation stage , select the time weighting factor , nonlinear mapping , . Substituting the set parameters into the associated calculation function, we can get: . Using the configured association calculation function, traverse the active control data and key quality parameters, substitute the active control data of the front process stage and the key quality parameters of the back process stage at different times into the function for calculation, obtain the association relationship between the various production process stages at different times, integrate these association relationships according to the event sequence, and establish a time-series production association.

[0041] The above calculation process, by introducing the concepts of time and weight, can more accurately reflect the relationship between various factors in the actual production process, and provide accurate correlation analysis means for subsequent abnormal propagation analysis and process optimization.

[0042] Furthermore, step S4 also includes: Step S41: Utilize the production feedback to perform optimization cost analysis and establish additional costs, wherein the additional costs include repair time costs and repair parameter adjustment costs; Step S42: using the additional cost to perform a repair-elimination balance analysis of the silicon carbide high-voltage MOSFET, and establishing a balance analysis result; Step S43: if the balance analysis result is an elimination result, generating an elimination optimization instruction; Step S44: marking the silicon carbide high-voltage MOSFET as unqualified according to the elimination optimization instruction, and ending the production feedback attention.

[0043] Specifically, additional costs are extra costs incurred during the production process. These costs are usually the costs that companies need to pay when there are abnormalities in the process or quality. Additional costs include but are not limited to repair time costs and repair parameter adjustment costs. Among them, repair time costs refer to the extra time consumed for repairs due to the discovery of problems. Repair parameter adjustment costs refer to the extra costs incurred by adjusting process parameters or operating procedures in order to repair problems found in production.

[0044] First, use the analysis results of production feedback to conduct cost analysis on the problems found. Use cost accounting software or spreadsheet tools to calculate the additional costs caused by abnormalities, such as the repair time caused by temperature abnormalities in the epitaxial growth stage, or the labor and material costs required to adjust the process parameters in the metallization stage, and then estimate the total cost required in the repair process. The repair time cost can be estimated by recording the time of downtime or process adjustment, while the repair parameter adjustment cost can be calculated by analyzing the equipment, materials and labor costs required to adjust the process parameters. By optimizing cost analysis, the additional costs in the repair process can be quantified, providing data support for subsequent repair-elimination balance decisions.

[0045] Using the additional cost data, combined with the product value, market price, expected benefits after repair of silicon carbide high-voltage MOSFET and other factors, a repair-elimination balance analysis is conducted. For example, by establishing a mathematical model, parameters such as additional cost, market value of the product, and expected added value after repair are included in the calculation. If the additional cost is too high, close to or exceeding the expected benefit after the product is repaired, then the balance analysis result may tend to eliminate; conversely, if the additional cost is relatively low and the product has a good expected benefit after repair, the balance analysis result may tend to repair. Performing a repair-elimination balance analysis can provide a decision-making basis for the product's handling method from the perspective of economy and benefits, avoiding the waste of resources or economic losses that may result from considering the product's handling method only from a technical perspective, and improving the rationality and economy of production decisions.

[0046] Elimination result is one of the balance analysis results, which means that after the repair-elimination balance analysis, it is determined that the product should be eliminated instead of repaired. Elimination optimization instructions are instructions generated after the product is determined to be an elimination result, which are used to guide operations such as marking the product as unqualified and stopping production feedback attention. If the balance analysis result is determined to be an elimination result, the elimination optimization instruction is automatically generated through the program algorithm.

[0047] According to the generated elimination optimization instructions, the corresponding silicon carbide high-voltage MOSFET products are marked in the production management system. This can be achieved through update operations in the database management system. For example, in the product database table, the status field of the product is updated to "unqualified" and the production feedback attention to the product is stopped. By marking the silicon carbide high-voltage MOSFET as unqualified and ending the production feedback attention, unqualified products in the production process can be cleared in time to avoid further investment in unqualified products. At the same time, it can also ensure that the resources of production feedback attention can be concentrated on valuable products, improving the efficiency and accuracy of the entire production management.

[0048] Further, step S42 includes: Step S421: performing a repair gain fitting on the silicon carbide high-voltage MOSFET to establish a repair gain fitting result; Step S422: Calculate the repair-elimination balance index using the repair benefit fitting result and the additional cost; Step S423: Establish the balance analysis result according to the repair-elimination balance index.

[0049] Specifically, by collecting historical production data, repair data and production feedback data, mathematical modeling methods (such as regression analysis, curve fitting, etc.) are used to fit the economic benefits brought by the repair, and the repair benefit fitting results are established. This repair benefit fitting result is a quantitative description of the profit situation of the silicon carbide high-voltage MOSFET after repair. It can be a numerical value, indicating the amount of profit expected to be obtained after the repair, or a profit function that can calculate the profit based on input variables such as different market conditions. For example, using statistical analysis software, based on historical market data, a linear regression model is established, with the performance improvement index after repair (such as the reduction in on-resistance) as the independent variable and the market price increase of the product as the dependent variable, and the repair benefit fitting results are obtained by fitting a large amount of data. Repair benefit fitting can provide production managers with the economic basis for repair decisions and help determine whether it is worth investing resources in repair. By quantifying the benefits brought by repair, production cost control and resource allocation can be optimized.

[0050] The repair-elimination balance index is a comprehensive index used to measure the economic benefit balance between repair and elimination. The index is derived from a comprehensive analysis of the repair benefit fitting results and additional costs, reflecting the relative advantages and disadvantages of repair and elimination under current conditions. Exemplarily, the calculation formula is set as: repair-elimination balance index = repair benefit fitting result / additional cost. The larger the value of the balance index, the higher the economic benefit of the repair and the lower the cost-effectiveness of the relative elimination; conversely, if the balance index is small, it may be more inclined to eliminate the product rather than repair. Through this index, a quantitative basis for repair or elimination can be provided for production management to avoid blind repair or elimination, thereby effectively saving production costs and improving resource utilization efficiency.

[0051] The balance analysis result is the final decision made based on the repair-elimination balance index, which clarifies whether to choose to repair or eliminate unqualified products. Based on the calculated repair-elimination balance index, the final result of the balance analysis is obtained according to the preset threshold or decision rule. For example, a threshold of the balance index is set. When the balance index is greater than the threshold, it is recommended to repair the product; when the balance index is less than the threshold, it is recommended to eliminate the product. Through the establishment of the balance analysis results, the production management system can make reasonable judgments at each decision node and choose the most appropriate action plan, thereby ensuring the efficiency of the production process and product quality while controlling costs.

[0052] Furthermore, step S5 further includes: Step S51: continuously record the production feedback concerns and establish a concern database; Step S52: evaluating the process stability of each production process stage according to the database of interest, and establishing stability feedback; Step S53: performing process optimization of the corresponding production process stage through the stable feedback.

[0053] Specifically, the attention database is a database dedicated to storing the content of continuous attention records, in which each record corresponds to the production feedback attention information at a certain moment in the production process, such as the value of a certain process parameter at a certain moment, the test results of product quality, etc. Use the database management system to establish an attention database. Continuously record the content of production feedback attention and store it in the attention database. By establishing an attention database, the effective storage of production feedback attention information is realized, so as to facilitate the tracing of various situations in the production process, provide a rich data source for subsequent operations such as process stability evaluation, and help to discover long-term trends and potential problems in the production process.

[0054] Extract data from each production process stage from the database of interest, analyze the stability of each process stage, and evaluate whether there are fluctuations, abnormalities or unstable factors affecting product quality. The analysis and evaluation process can use statistical analysis methods such as standard deviation, coefficient of variation, control charts, etc. to establish an indicator related to the number and frequency of abnormal data to reflect the stability of the process stage.

[0055] According to the results of stable feedback, the problem is analyzed in depth, and then the corresponding process parameters, operating steps or equipment settings are adjusted or improved to improve the stability and efficiency of the production process and improve product quality. For example, if the ion implantation and diffusion stages are unstable, by checking the detailed data in the database, it is found that the dose of the implanted ions fluctuates greatly. Then the parameter settings of the ion implanter can be adjusted, or the stability of the ion implantation equipment can be checked. If it is an equipment problem, the equipment can be repaired or upgraded. For the adjustment of process parameters, automated control systems such as programmable logic controllers (PLCs) can be used to accurately control process parameters.

[0056] Furthermore, the method described in the embodiment of the present application also includes: Step S61: focusing on configuring quality inspection feedback according to the production feedback; Step S62: Perform a focus quality inspection on the corresponding silicon carbide high-voltage MOSFET finished product through the quality inspection feedback, and manage the silicon carbide high-voltage MOSFET according to the focus quality inspection result.

[0057] Specifically, quality inspection feedback is a feedback mechanism configured according to production feedback concerns, which contains various information related to the quality inspection of finished silicon carbide high-voltage MOSFET products, such as key inspection indicators, inspection methods, inspection frequency, etc. Analyze the data in the production feedback concerns, determine the key factors that may affect product quality in the production process, and configure quality inspection feedback according to these key factors. For example, the production feedback concerns show that the ion implantation and diffusion stages have a greater impact on product quality. For example, the accuracy of the implantation dose will seriously affect the electrical properties of the silicon carbide high-voltage MOSFET. In the quality inspection feedback, the detection of the implantation dose is used as a key indicator. By configuring quality inspection feedback according to production feedback concerns, we can focus on the key factors that affect product quality in the production process, make quality inspection more targeted, and improve the efficiency and effectiveness of quality inspection.

[0058] Focus on quality inspection is a kind of finished product quality inspection of silicon carbide high-voltage MOSFET based on quality inspection feedback. It is different from conventional comprehensive quality inspection, but focuses on the indicators, methods and frequency set in the quality inspection feedback.

[0059] According to the inspection content and requirements in the configured quality inspection feedback, the quality of the finished silicon carbide high-voltage MOSFET is specifically inspected using automated quality inspection instruments, online sensors, test benches, etc. For example, if a temperature fluctuation concern is found in the epitaxial growth stage during the production process, a temperature sensitivity test is performed at the finished product stage, or the thickness and uniformity of the epitaxial layer is detected. Subsequent management of the silicon carbide high-voltage MOSFET is performed based on the quality inspection results of concern, such as whether to repair, rework, eliminate, or continue to the next step of production. For example, if the test results are qualified, the product is put into storage, packaged, and shipped. If the test results are unqualified, the product is marked as unqualified in the ERP system, isolated and stored, and the reasons for the unqualified are analyzed. If there may be a problem in a certain link in the production process, the relevant information in the production feedback concern can be viewed, and then corresponding treatment measures are taken according to the cause, such as reworking or scrapping the product.

[0060] Through quality inspection feedback, we can accurately monitor problems in each production link that may affect the quality of the final product, promptly detect and handle unqualified products, adjust the production process, reduce scrap rate, ensure that product quality meets standards, and optimize overall production efficiency.

[0061] In summary, the process dynamic control method for silicon carbide high-voltage MOSFET provided in the embodiment of the present application has the following technical effects: Overall, the embodiment of the present application creates an abnormal propagation network through a comprehensive analysis of the silicon carbide high-voltage MOSFET production process, establishes an intelligent management platform to communicate with the edge sensor group, collects production data in real time for abnormal propagation analysis, and realizes dynamic optimization of the entire production process and real-time abnormality warning, thereby improving the production efficiency and product quality of silicon carbide high-voltage MOSFET, and enhancing the performance and reliability of the product in harsh working environments such as high voltage, high frequency, and high temperature.

[0062] Embodiment 2, as Figure 3 As shown, an embodiment of the present application provides a process dynamic control device for silicon carbide high-voltage MOSFET, the device comprising: The production process analysis module 10 is used to analyze the production process of silicon carbide high-voltage MOSFET and establish production process stages, which include pre-processing stage, epitaxial growth stage, ion implantation and diffusion stage, metallization and contact structure formation stage, and packaging and testing stage.

[0063] The historical production data analysis module 20 is used to call the historical production data set using the production process stage, perform time series correlation analysis based on the historical production data set, and create an abnormal propagation network using the time series correlation analysis results.

[0064] The management platform construction module 30 is used to establish an intelligent management platform based on the abnormal propagation network. The intelligent management platform digitally communicates with the edge sensor group distributed in each production process stage. The edge sensor is used to collect production data of each production process stage.

[0065] The abnormal propagation analysis module 40 is used to use the intelligent management platform to read the real-time production data of the edge sensor, perform abnormal propagation analysis based on the abnormal propagation network, and establish production feedback attention.

[0066] The process dynamic optimization module 50 is used to perform process dynamic optimization management of the production process stage through the production feedback attention.

[0067] Furthermore, the historical production data analysis module 20 of the embodiment of the present application is also used to perform the following steps: After mapping the historical production data set to the production process stage, the historical production data set is segmented to establish active control data and key quality parameters, wherein the active control data includes temperature, pressure, time, and gas flow, and the key quality parameters include epitaxial growth rate and heat dissipation layer quality; the autocorrelation channel and the cross-correlation channel are called to perform time lag relationship analysis between the active control data and the key quality parameters, and establish a timing lag association within the production process stage; an association analysis channel is established, and the association analysis channel is used to perform association analysis between production process stages, and a timing production association is established; and a timing association analysis result is established using the timing lag association and the timing production association.

[0068] Furthermore, the historical production data analysis module 20 of the embodiment of the present application is also used to perform the following steps: Configure the correlation calculation function of the correlation analysis channel as follows: ;in, Characterization process stage and process stages The association, Represents the time point index, which is used to indicate the time sequence of data. is the total length of the time series, representing the total number of data points involved in the association calculation, is the time weighting factor, characterizing each time point Since the production process is carried out over time, data at different time points may have different impacts on the final correlation results. The time weighting factor is used to adjust the degree of this impact. Characterize active control data of the front-end process stage, Characterize key quality parameters of post-process stages, A nonlinear mapping function that characterizes the active control data of the preceding process stage, Nonlinear mapping functions that characterize key quality parameters of post-process stages, The square term of the nonlinear mapping function that characterizes the active control data, A square term of a nonlinear mapping function characterizing a key quality parameter; performing association analysis using the association calculation function to establish the time series production association.

[0069] Furthermore, the abnormal propagation analysis module 40 of the embodiment of the present application is also used to perform the following steps: The production feedback concern is used to perform an optimization cost analysis to establish additional costs, wherein the additional costs include repair time costs and repair parameter adjustment costs; the additional costs are used to perform a repair-elimination balance analysis of the silicon carbide high-voltage MOSFET to establish a balance analysis result; if the balance analysis result is an elimination result, an elimination optimization instruction is generated; according to the elimination optimization instruction, the silicon carbide high-voltage MOSFET is marked as unqualified, and the production feedback concern is ended.

[0070] Furthermore, the abnormal propagation analysis module 40 of the embodiment of the present application is also used to perform the following steps: Performing repair benefit fitting on the silicon carbide high-voltage MOSFET to establish a repair benefit fitting result; calculating a repair-elimination balance index using the repair benefit fitting result and the additional cost; and establishing the balance analysis result based on the repair-elimination balance index.

[0071] Furthermore, the process dynamic optimization module 50 of the embodiment of the present application is also used to perform the following steps: The production feedback attention is continuously recorded to establish an attention database; the process stability of each production process stage is evaluated based on the attention database to establish stable feedback; and the process of the corresponding production process stage is optimized through the stable feedback.

[0072] Furthermore, the apparatus described in the embodiment of the present application further includes a quality feedback module, and the quality feedback module is used to perform the following steps: Based on the production feedback, quality inspection feedback is configured; quality inspection of the corresponding silicon carbide high-voltage MOSFET finished products is performed through the quality inspection feedback, and silicon carbide high-voltage MOSFET management is performed based on the quality inspection results.

[0073] Through the above-mentioned detailed description of the process dynamic control method for silicon carbide high-voltage MOSFET in this specification, those skilled in the art can clearly understand the process dynamic control device for silicon carbide high-voltage MOSFET in this embodiment. For the device disclosed in Example 2, since it corresponds to the method disclosed in Example 1 and has corresponding functional modules and beneficial effects, the relevant parts can be referred to the description of the method part.

[0074] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A process dynamic control method for silicon carbide high voltage MOSFET, characterized in that: The method comprises: Analyze the production process of silicon carbide high-voltage MOSFET and establish the production process stages, which include pre-processing stage, epitaxial growth stage, ion implantation and diffusion stage, metallization and contact structure formation stage, and packaging and testing stage; Using the production process stage to call a historical production data set, performing a time series correlation analysis based on the historical production data set, and creating an abnormal propagation network using the time series correlation analysis result; Establishing an intelligent management platform based on the abnormal propagation network, wherein the intelligent management platform digitally communicates with an edge sensor group distributed at each production process stage, wherein the edge sensor is used to collect production data at each production process stage; After using the intelligent management platform to read the real-time production data of the edge sensor, an abnormal propagation analysis is performed based on the abnormal propagation network to establish production feedback attention; The production feedback is used to focus on dynamic optimization management of the process in the production process stage.

2. The process dynamic control method for silicon carbide high voltage MOSFET according to claim 1, characterized in that: The performing of time series correlation analysis based on the historical production data set and creating an abnormal propagation network using the time series correlation analysis result includes: After mapping the production process stages of the historical production data set, data segmentation is performed on the historical production data set to establish active control data and key quality parameters, wherein the active control data includes temperature, pressure, time, and gas flow, and the key quality parameters include epitaxial growth rate and heat dissipation layer quality; Calling the autocorrelation channel and the cross-correlation channel to analyze the time lag relationship between the active control data and the key quality parameters, and establishing the timing lag association within the production process stage; Establishing a correlation analysis channel, using the correlation analysis channel to perform correlation analysis between production process stages, and establishing a time-series production correlation; The timing lag correlation and the timing production correlation are used to establish a timing correlation analysis result.

3. The process dynamic control method for silicon carbide high voltage MOSFET according to claim 2, characterized in that: The using the correlation analysis channel to perform correlation analysis between production process stages and establish time-series production correlation includes: Configure the correlation calculation function of the correlation analysis channel as follows: ； in, Characterization process stage and process stages The association, Represents the time point index, which is used to indicate the time sequence of data. is the total length of the time series, representing the total number of data points involved in the association calculation, is the time weighting factor, characterizing each time point The weight of Characterize active control data of the front-end process stage, Characterize key quality parameters of post-process stages, A nonlinear mapping function that characterizes the active control data of the preceding process stage, Nonlinear mapping functions that characterize key quality parameters of post-process stages, The square term of the nonlinear mapping function that characterizes the active control data, A square term of a nonlinear mapping function characterizing a key quality parameter; performing association analysis using the association calculation function to establish the time series production association.

4. The process dynamic control method for silicon carbide high voltage MOSFET according to claim 1, characterized in that: After using the intelligent management platform to read the real-time production data of the edge sensor, an abnormal propagation analysis is performed based on the abnormal propagation network to establish production feedback attention, including: Utilizing the production feedback to perform optimization cost analysis and establish additional costs, wherein the additional costs include repair time costs and repair parameter adjustment costs; Using the additional cost to perform a repair-elimination balance analysis of the silicon carbide high-voltage MOSFET, and establishing a balance analysis result; If the balance analysis result is an elimination result, generating an elimination optimization instruction; The silicon carbide high voltage MOSFET is marked as unqualified according to the elimination optimization instruction, and the production feedback attention is ended.

5. The process dynamic control method for silicon carbide high voltage MOSFET according to claim 4, characterized in that: The method of using the additional cost to perform a repair-elimination balance analysis of the silicon carbide high-voltage MOSFET and establishing a balance analysis result includes: Performing a repair gain fitting on the silicon carbide high-voltage MOSFET to establish a repair gain fitting result; Calculating a repair-elimination balance index using the repair benefit fitting result and the additional cost; The balance analysis result is established according to the repair-elimination balance index.

6. The process dynamic control method for silicon carbide high voltage MOSFET according to claim 1, characterized in that: The method further comprises: Continuously record the production feedback concerns and establish a concern database; Conducting process stability evaluation at each production process stage based on the database of interest and establishing stability feedback; The stable feedback is used to perform process optimization at the corresponding production process stage.

7. The process dynamic control method for silicon carbide high voltage MOSFET according to claim 1, characterized in that: The method further comprises: Focus on configuring quality inspection feedback based on said production feedback; The quality inspection feedback is used to perform quality inspection on the corresponding silicon carbide high-voltage MOSFET finished product, and the silicon carbide high-voltage MOSFET is managed according to the quality inspection result.

8. A process dynamic control device for silicon carbide high voltage MOSFET, characterized in that: The device is used to execute the process dynamic control method for silicon carbide high-voltage MOSFET according to any one of claims 1 to 7, comprising: The production process analysis module is used to analyze the production process of silicon carbide high-voltage MOSFET and establish the production process stages, which include pre-processing stage, epitaxial growth stage, ion implantation and diffusion stage, metallization and contact structure formation stage, and packaging and testing stage; A historical production data analysis module, used to call the historical production data set using the production process stage, perform time series correlation analysis based on the historical production data set, and create an abnormal propagation network using the time series correlation analysis result; A management platform building module, used to establish an intelligent management platform based on the abnormal propagation network, the intelligent management platform digitally communicates with an edge sensor group distributed at each production process stage, the edge sensor is used to collect production data at each production process stage; An abnormal propagation analysis module is used to use the intelligent management platform to read the real-time production data of the edge sensor, perform abnormal propagation analysis based on the abnormal propagation network, and establish production feedback attention; The process dynamic optimization module is used to carry out process dynamic optimization management of the production process stage through the production feedback attention.

Images