Method and system for monitoring abnormal circulation of silicon wafers in track and storage medium

Abstract

The invention relates to the technical field of silicon wafer circulation monitoring, in particular to a method and system for monitoring abnormal circulation of silicon wafers in a track and a storage medium. The monitoring method comprises the steps that a large belt runner and a small belt line are provided, the large belt runner is provided with a plurality of storage areas P1, P2, P3,..., Pn used for storing silicon wafers, and each storage area is provided with an optical sensor used for detecting the existence state and the circulation condition of the silicon wafers in real time; and the silicon wafer is transferred between the flower basket and the large belt runner through the X / Y / Z transplanting module, so that the silicon wafer can accurately flow from one process stage to another process stage. According to the invention, by introducing the logic processing unit and the optimized alarm processing method, the abnormity monitoring efficiency of the silicon wafer circulation process in the CVD equipment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of silicon wafer circulation monitoring, and in particular to a method, system and storage medium for monitoring abnormal silicon wafer circulation in a track. Background Art

[0002] In chemical vapor deposition (CVD) equipment, silicon wafers are circulated through automated equipment to remove and transfer silicon wafers from and to the baskets. Traditional monitoring systems rely solely on sensors to detect the presence of silicon wafers, and are unable to effectively handle repeated alarms caused by position changes during the transfer process. This not only increases the workload of operators, but can also lead to false alarms and missed alarms, affecting the normal operation efficiency and production quality of the equipment. Summary of the invention

[0003] The present invention provides a method, system and storage medium for monitoring abnormal flow of silicon wafers in a track to solve the problem of repeated alarms caused by position changes of abnormal silicon wafers during transmission.

[0004] The technical solution of the present invention to solve the above technical problems is as follows:

[0005] On the one hand, a method for monitoring abnormal flow of silicon wafers in a track is provided, which is applied to the alarm processing of missing or damaged silicon wafers in the flow channel, and the monitoring method comprises:

[0006] A large belt flow channel and a small belt line are provided, wherein the large belt flow channel is provided with a plurality of storage areas P 1 , P 2 , P 3 ,…,P n , used to store the silicon wafers, each of the storage areas is provided with an optical sensor for real-time detection of the existence and circulation of the silicon wafers;

[0007] The silicon wafers are transferred between the flower basket and the large belt flow channel by an X / Y / Z transfer module to ensure that the silicon wafers can be accurately transferred from one process stage to another;

[0008] During the silicon wafer transmission process, real-time monitoring of the status information of the silicon wafer in each storage area, including the location, access status and physical status;

[0009] When it is detected that the silicon wafer in a certain storage area is missing or damaged, an initial alarm signal is generated, the initial alarm signal is transmitted to a logic processing unit, an alarm process is performed, and a subsequent transmission operation of the silicon wafer is adjusted;

[0010] During the transmission process, the status information of the silicon wafers in each storage area is continuously updated, and the alarm logic is adjusted according to the updated status information.

[0011] Furthermore, the steps of the alarm processing include:

[0012] Recording the storage area where the initial alarm signal occurs and the unique identification information of the silicon wafer, including the batch number and the process flow section number;

[0013] Analyze the transmission path of the silicon wafer, and determine whether the current alarm is a repeated alarm by combining the built-in path algorithm with the historical alarm records. The path algorithm stores the flow direction of the silicon wafer in a tree structure, and each node represents a storage area. When an abnormal state of a node is detected, the path algorithm traces the state of the upstream and downstream nodes of the node to avoid repeated alarms due to position changes during the transmission of the silicon wafer;

[0014] According to the analysis results, the logic processing unit generates a processing alarm signal and an operation flow;

[0015] The processed alarm signal is transmitted to the control unit to prompt the operator to confirm and operate.

[0016] Furthermore, the step of the optical sensor detecting the state of the silicon wafer includes:

[0017] One or more optical sensors are arranged along the large belt flow path, each optical sensor group includes at least two photosensitive elements at different angles, and is used to obtain multi-angle image data of the silicon wafer;

[0018] The acquired image data is preprocessed through the image processing module, including grayscale and filtering steps, to eliminate ambient light interference and improve detection accuracy;

[0019] A deep learning algorithm is used to analyze the preprocessed image data to determine whether the storage area contains silicon wafers. If so, it is further determined whether the silicon wafers have signs of damage. The deep learning algorithm uses a convolutional neural network model and is trained using a large amount of pre-labeled silicon wafer image data to achieve high-precision silicon wafer status recognition.

[0020] Furthermore, the deep learning algorithm also includes:

[0021] The convolutional neural network model is pre-trained using a large amount of image data from the semiconductor processing technology, and then fine-tuned using a small amount of image data from the CVD process to accelerate the convergence speed of the model;

[0022] A mechanism for dynamically adjusting the learning rate of the neural network is adopted to dynamically adjust the learning rate according to the current training error. The adjustment calculation formula of the learning rate is:

[0023] Among them, α is the learning rate, α0 is the initial learning rate, α min is the minimum learning rate, T is the preset maximum number of training times, t is the current number of training times, and λ is the exponential decay speed parameter.

[0024] Furthermore, before generating the processing alarm signal, the logic processing unit predicts the state of the next storage area of ​​the silicon wafer through the state prediction module using the Kalman filter, and the state prediction module predicts the future position state of the silicon wafer in real time based on the current position variable, speed variable and acceleration variable of the silicon wafer, combined with the historical transmission data and model parameters;

[0025] By comparing the predicted results with the actual detection results, it is determined whether it is a sensor false alarm or a silicon chip abnormality, and the alarm signal generation logic is adjusted according to the judgment result to reduce the occurrence of false alarms.

[0026] Furthermore, after prompting the operator to confirm, the monitoring method further comprises the following steps:

[0027] The interactive interface displays alarm information and wafer status in real time, including the specific location, time, and degree of wafer damage, so that operators can quickly locate the problem.

[0028] If the operator determines that the current alarm is a false alarm, the false alarm information is transmitted back to the logic processing unit through a feedback mechanism for learning and optimization to reduce the probability of future false alarms;

[0029] If the operator confirms that the silicon wafer is missing or damaged, the exception handling process is initiated.

[0030] Furthermore, the specific steps of continuing to perform the silicon wafer transfer operation according to the operator's confirmation instruction include:

[0031] Once the operator confirms the alarm, the current silicon wafer transfer process is suspended to avoid losses due to abnormalities;

[0032] The control unit adjusts the transmission path of subsequent silicon wafers according to the confirmed abnormal situation, including calculating the optimal transmission order of the remaining silicon wafers and giving priority to silicon wafers with longer transmission distances and lower risks, thereby reducing equipment wear caused by repeated operations in the same area;

[0033] By controlling the action of the X / Y / Z transplanting module, the transfer of the silicon wafers between the large belt flow channel and the flower basket is adjusted until all abnormal situations are handled.

[0034] Furthermore, the monitoring method further comprises:

[0035] By comprehensively analyzing the alarm records within a fixed time window, the fuzzy logic algorithm is used to evaluate the abnormal credibility. The calculation formula of the fuzzy logic algorithm is:

[0036] If C>T, it is considered that there is a high-confidence abnormality in the window, where n 1 ,n 2 ,…,n k is the number of alarms in different storage areas within the time window, w 1 ,w 2 ,…,w k is the corresponding weight, T is the abnormality determination threshold;

[0037] The fusion result of all the alarm data within the time window serves as the basis for the processing unit to generate an alarm signal.

[0038] On the other hand, a system for monitoring abnormal flow of silicon wafers in a track is provided, the monitoring system comprising:

[0039] A large belt runner and a small belt line, wherein the large belt runner is provided with a plurality of storage areas P 1 , P 2 , P 3 ,…,P n , used to store the silicon wafers, each of the storage areas is equipped with an optical sensor for real-time monitoring of the status of the silicon wafers;

[0040] An X / Y / Z transfer module, used to transfer the silicon wafers between the flower basket and the large belt flow channel, wherein the transfer module uses a vacuum adsorption method to ensure that the silicon wafers do not slip during the transfer process;

[0041] A logic processing unit, used to receive the wafer status detection information, and process the alarm information by the above-mentioned method for monitoring the abnormal flow of wafers in the track, so as to avoid repeated alarms caused by position changes;

[0042] A control unit is used to receive the processing alarm signal and prompt the operator to confirm, and perform subsequent silicon wafer flow operations according to the operator's feedback

[0043] An interactive interface that displays abnormal monitoring data and operating instructions in real time, assisting operators in responding to and handling abnormal situations;

[0044] An exception handling process module starts the corresponding exception handling process according to the result of operation confirmation, including automatically recording exception information and arranging maintenance operations.

[0045] On the other hand, a readable storage medium is provided, on which a program for monitoring abnormal flow of silicon wafers in a track is stored. When the monitoring program is executed by a processor, the processor executes the steps of the method for monitoring abnormal flow of silicon wafers in a track as described above.

[0046] The beneficial effects of the present invention are:

[0047] The present invention improves the abnormal monitoring efficiency of the silicon wafer circulation process in the CVD equipment by introducing a logic processing unit and an optimized alarm processing method. Specifically, by real-time monitoring of the silicon wafer status of each storage area, and combining the transmission path of the silicon wafer and the historical alarm record, the alarm signal is logically processed, effectively avoiding the problem of repeated alarms caused by changes in the position of the silicon wafer. This processing method not only reduces the repeated confirmation work of the operator and improves work efficiency, but also reduces the risk of false alarms and missed alarms, ensuring that the equipment can respond promptly and accurately when facing abnormal situations. In addition, the present invention can also continuously update the status information of each storage area during the silicon wafer transmission process, and dynamically adjust the alarm logic according to this information, further improving the reliability of the system, and providing a solution for silicon wafer monitoring in the semiconductor production process.

[0048] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of a silicon wafer flower basket in one embodiment of the present invention Figure 1 ;

[0050] Figure 2 This is a schematic diagram of a silicon wafer flower basket in one embodiment of the present invention;

[0051] Figure 3 Schematic diagram of silicon wafers entering a basket in one embodiment of the present invention Figure 1 ;

[0052] Figure 4 Schematic diagram of silicon wafers entering a basket in one embodiment of the present invention Figure 2 ;

[0053] Figure 5 This is a schematic diagram of the process of silicon wafers discharging a flower basket in one embodiment of the present invention;

[0054] Figure 6 This is a schematic diagram of the process of silicon wafers entering the flower basket in one embodiment of the present invention;

[0055] Figure 7 The figure is a flow chart of a method for monitoring abnormal flow of silicon wafers in a track according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] The term "comprise" and any variation thereof in the specification and claims of the present application are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. In addition, the use of "and / or" in the specification and claims to indicate at least one of the connected objects, such as A and / or B, means including A alone, B alone, and A and B.

[0058] In the embodiments of the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0059] The present invention provides the following preferred embodiments:

[0060] Embodiment 1

[0061] In order to solve the problem of repeated alarms caused by position changes during the circulation of silicon wafers in CVD equipment, this embodiment proposes a method for monitoring abnormal circulation of silicon wafers in a track. Specifically, in this embodiment, a large belt flow channel containing multiple storage areas and a small belt line are used, and each storage area is equipped with an optical sensor for real-time detection of the status of silicon wafers. The X / Y / Z transplanting module is responsible for taking the silicon wafers out of the flower basket and placing them on the large belt flow channel, or transferring them back to the flower basket from the large belt flow channel. Figure 1 and Figure 2 Demonstrate the process of transferring silicon wafers from the small belt to the large belt in the flower basket. Figure 1 If silicon wafer B is missing from P1, a silicon wafer alarm is generated. After confirmation by personnel, enter Figure 2 : The small belt is transmitted to P1, the P1 silicon wafer is transmitted to P2, P2 is transmitted to P3, P3 is transmitted to P4, and no new alarm is generated due to the position change of silicon wafer B. If the remaining silicon wafers are not detected during the transmission process, new alarms can still be generated. See the process. Figure 5 .and Figure 3and Figure 4 Demonstrate the process of transferring silicon wafers from the large belt to the small belt. First, the X / Y / Z transfer module places the silicon wafer on the large belt. Figure 3 , Figure 3 A silicon wafer is missing at P3, generating a silicon wafer alarm. After confirmation by personnel, enter Figure 4 : P4 silicon wafer is transmitted to P3, P3 is transmitted to P2, ..., P1 is transmitted to the small belt, and no new alarm is generated due to the position change of silicon wafer B. If the remaining silicon wafers are not detected during the transmission process, new alarms can still be generated. See the process. Figure 6 This design not only ensures the accurate transfer of silicon wafers at each process stage, but also effectively avoids repeated alarms through real-time monitoring and the cooperation of the logic processing unit.

[0062] like Figure 7 As shown, the monitoring method comprises the following steps:

[0063] S100: Provides one large belt runner and one small belt line, the large belt runner has multiple storage areas P 1 , P 2 , P 3 ,…,P n , used to store silicon wafers. Each storage area is equipped with an optical sensor to detect the existence and flow of silicon wafers in real time.

[0064] S200: The X / Y / Z transfer module is used to transfer the silicon wafers between the flower basket and the large belt flow channel to ensure that the silicon wafers can flow accurately from one process stage to another.

[0065] S300: During the silicon wafer transmission process, the status information of the silicon wafers in each storage area is monitored in real time, including the location, access status and physical status.

[0066] S400: When it is detected that a silicon wafer in a certain storage area is missing or damaged, an initial alarm signal is generated, the initial alarm signal is transmitted to the logic processing unit, an alarm process is performed, and a subsequent silicon wafer transmission operation is adjusted.

[0067] S500: During the transmission process, the status information of the silicon wafers in each storage area is continuously updated, and the alarm logic is adjusted according to the updated status information.

[0068] During the wafer transfer process, optical sensors detect the status of the wafers in each storage area in real time. These sensors can use high-resolution industrial cameras to capture multi-angle images and detailed status information of the wafers. It should be understood that the selection of optical sensors should take into account their detection accuracy and response speed to adapt to the high-speed production environment. The data collected by each sensor is transmitted to the logic processing unit through a wired or wireless communication module for real-time analysis and processing.

[0069] Furthermore, when the logic processing unit receives the initial alarm signal, it first records the storage area where the signal occurs and the identification information of the silicon wafer. The identification information may include the batch number of the silicon wafer, the process flow section number, etc., for subsequent tracking and processing. Secondly, the logic processing unit determines whether the current alarm is a repeated alarm based on the transmission path of the silicon wafer and the historical alarm records. This judgment process is implemented through a path algorithm, which uses a tree structure to store the transmission path of the silicon wafer, and each node represents a storage area. When an abnormal node state is detected, the algorithm traces the upstream and downstream node states of the node to confirm whether it is a repeated alarm of the same silicon wafer in the same process flow. Through this multi-node state analysis, false alarms and repeated alarms caused by position changes can be effectively reduced.

[0070] Furthermore, after the logic processing unit generates a processing alarm signal, the signal is transmitted to the control unit. The control unit displays the alarm information and silicon wafer status in real time through the interactive interface, assisting the operator to quickly locate the problem and confirm it. It should be understood that the design of the interactive interface should be intuitive and efficient, and can quickly display key information such as the alarm location, time, and degree of damage to the silicon wafer. After the operator confirms, the control unit adjusts the subsequent silicon wafer transmission operation according to the confirmation instruction to ensure the continuity and safety of the production process. For example, if it is confirmed that there is a missing or damaged silicon wafer, the control unit will immediately suspend the current transmission process and rearrange the transmission path of the subsequent silicon wafer to avoid equipment wear due to repeated operations at the same location.

[0071] Furthermore, during the silicon wafer transmission process, the logic processing unit continuously updates the silicon wafer status information of each storage area and dynamically adjusts the alarm logic based on this information. For example, when the sensor detects that the silicon wafer status in a storage area has changed, the logic processing unit will recalculate the alarm threshold of the area in combination with the current time window and transmission path. This process is implemented through a state prediction module based on a Kalman filter. The state prediction module predicts the future position state of the silicon wafer based on the current position, speed and acceleration of the silicon wafer, combined with historical transmission data and model parameters. By comparing the prediction results with the actual detection results, it is determined whether there are false alarms or repeated alarms, thereby optimizing the alarm processing process.

[0072] The benefit of this embodiment is that, through the cooperation of the logic processing unit and the state prediction module, the problem of repeated alarms caused by position changes during the circulation of silicon wafers is effectively solved. This method not only reduces the repeated confirmation work of operators and improves work efficiency, but also reduces the risk of false alarms and missed alarms, ensuring that the equipment can respond promptly and accurately when facing abnormal situations. In addition, by continuously updating silicon wafer status information and dynamically adjusting alarm logic, the system can respond to various complex production environments more intelligently.

[0073] Embodiment 2

[0074] In order to solve the problem of repeated alarms caused by position changes during the transmission of silicon wafers, this embodiment further optimizes the alarm processing steps. Specifically, when the optical sensor detects that there is a missing or damaged silicon wafer in a storage area, an initial alarm signal is generated. The signal is transmitted to the logic processing unit, which first records the storage area where the initial alarm signal occurs and the unique identification information of the silicon wafer, including the batch number and process flow section number of the silicon wafer. Doing so not only helps to track the source of the silicon wafer, but also quickly locates the flow path of the silicon wafer in the historical records.

[0075] Furthermore, the logic processing unit uses a built-in path algorithm combined with historical alarm records to determine whether the current alarm is a repeated alarm. The path algorithm uses a tree structure to store the flow of silicon wafers, where each node represents a storage area. When an abnormal node state is detected, the path algorithm traces the upstream and downstream node states of the node to determine whether the same silicon wafer has continuous alarms at multiple locations. It should be understood that this method can effectively avoid repeated alarms caused by multiple movements of silicon wafers in the same process flow, and improve the accuracy of system alarms.

[0076] Further, based on the analysis results, the logic processing unit generates processed alarm signals and corresponding operation procedures. These signals include whether it is confirmed as a new alarm and subsequent processing steps, such as stopping the transmission line or notifying maintenance personnel. The processed alarm signal is transmitted to the control unit, which prompts the operator to confirm and operate through the interactive interface.

[0077] Furthermore, after the operator confirms, the system will continue to perform subsequent silicon wafer transmission operations and continuously update the silicon wafer status information of each storage area during the transmission process. According to the latest status information, the logic processing unit will dynamically adjust the alarm logic to ensure the timeliness and accuracy of alarm processing.

[0078] The benefit of this embodiment is that, through the intelligent analysis and processing of the alarm signal by the logic processing unit, the repeated alarms caused by the change of the silicon wafer position are effectively reduced, the alarm accuracy of the system and the operating efficiency of the equipment are improved, and the workload of the operators is reduced.

[0079] Embodiment 3

[0080] In order to solve the problem of detecting the state of silicon wafers during the transmission process, the method of optical sensors detecting the state of silicon wafers is further refined. Specifically, in this embodiment, multiple groups of optical sensors are arranged on the upper edge of the large belt flow channel, and each group of optical sensors contains at least two photosensitive elements at different angles, which are used to obtain multi-angle image data of silicon wafers. These photosensitive elements can be high-resolution industrial cameras to ensure the accuracy and reliability of the detection results. Multi-angle image data can more comprehensively reflect the state of silicon wafers, including position, integrity and surface condition.

[0081] Furthermore, the acquired multi-angle image data will be pre-processed by the image processing module, including grayscale and filtering steps. Grayscale can convert color images into grayscale images, simplify the complexity of data processing, and reduce the computational burden. The filtering step is used to eliminate the interference of ambient light, improve image quality, and ensure that subsequent analysis is more accurate. It can be understood that these pre-processing steps are crucial to improving detection accuracy.

[0082] Furthermore, the pre-processed image data is then input into a deep learning algorithm based on a convolutional neural network (CNN) model, which is trained using a large amount of pre-labeled silicon wafer image data to achieve high-precision silicon wafer status recognition. Convolutional neural networks can automatically learn various features of silicon wafers, such as edges, textures, and contours, so that during detection, they can quickly and accurately determine whether the storage area contains silicon wafers and whether the silicon wafers show signs of damage. It should be understood that the introduction of deep learning algorithms has greatly improved the automation level and accuracy of status detection.

[0083] The benefit of this embodiment is that, through the acquisition of multi-angle images and advanced image processing and deep learning algorithms, the accuracy and reliability of silicon wafer status detection are significantly improved, production interruptions due to false detection are reduced, and the operating efficiency and production quality of the equipment are improved.

[0084] Embodiment 4

[0085] In order to solve the convergence speed problem in the deep learning model training process, this embodiment further optimizes the training method of the convolutional neural network (CNN) model. Specifically, by pre-training the CNN model using a large amount of image data in the semiconductor processing technology, it can learn the basic characteristics of the silicon wafer. Then, a small amount of image data in the CVD process is used for fine-tuning to further adapt to the current application scenario. This method is called transfer learning, which can effectively reduce training time and improve the convergence speed of the model.

[0086] Furthermore, this embodiment introduces a mechanism for dynamically adjusting the learning rate of the neural network. The learning rate adjustment calculation formula is:

[0087]

[0088] Among them, α is the learning rate, α 0 is the initial learning rate, α min is the minimum learning rate, T is the preset maximum number of training times, t is the current number of training times, and λ is the exponential decay rate parameter. This dynamic adjustment mechanism can automatically adjust the learning rate according to the changes in the current training error, so that the model can quickly learn the initial parameters in the early stage of training, and can more finely optimize the model in the later stage of training to improve the convergence accuracy.

[0089] It is understandable that through the methods of pre-training and fine-tuning, the model can adapt to specific application scenarios more quickly and reduce training time. At the same time, the mechanism of dynamically adjusting the learning rate makes the model training process more flexible, improving the training efficiency and model performance.

[0090] The benefit of this embodiment is that, through the method of transfer learning and dynamic adjustment of learning rate, the training time of the model is shortened, the convergence speed and performance of the model are improved, so that the silicon wafer status detection system can be put into practical application more quickly.

[0091] Embodiment 5

[0092] In order to solve the problem of false alarms caused by sensor false alarms or silicon wafer abnormalities during the transmission process of the silicon wafer, this embodiment further optimizes the alarm processing method of the logic processing unit. Specifically, before generating and processing the alarm signal, the logic processing unit uses the Kalman filter through the state prediction module to predict the state of the next storage area of ​​the silicon wafer. The state prediction module predicts the future position state of the silicon wafer in real time based on the current position, speed and acceleration variables of the silicon wafer, combined with historical transmission data and model parameters.

[0093] Furthermore, by comparing the prediction results with the actual detection results, the logic processing unit can determine whether the current alarm is a false alarm of the sensor or there is indeed an abnormality in the silicon chip. The Kalman filter can effectively process measurement data containing noise and provide accurate prediction results. When the prediction result is consistent with the actual detection result, it means that the sensor detection is accurate and the current alarm is a valid alarm; when the two are inconsistent, it means that the sensor may be falsely alarmed or the silicon chip has not yet reached the target position. It should be understood that this method of combining prediction and detection can significantly reduce the occurrence of false alarms and improve the alarm accuracy and reliability of the system.

[0094] Furthermore, based on the judgment result, the logic processing unit will adjust the alarm signal generation logic. If it is confirmed that the sensor is falsely alarmed, the logic processing unit will suppress the alarm signal to avoid unnecessary interference; if it is confirmed that the silicon chip is indeed abnormal, the logic processing unit will generate an alarm signal and notify the control unit to prompt the operator to confirm and handle it.

[0095] The benefit of this embodiment is that, by introducing the state prediction module and the Kalman filter, false alarms caused by sensor false alarms or the silicon chip not reaching the target position are effectively avoided, thereby improving the alarm accuracy of the system.

[0096] Embodiment 6

[0097] In order to solve the feedback problem of operators when confirming the alarm, this embodiment further optimizes the operation process after the alarm processing. Specifically, when the logic processing unit generates a processed alarm signal, the signal will be transmitted to the control unit, and the alarm information and silicon chip status will be displayed in real time through the interactive interface. The alarm information includes detailed information such as the specific alarm location, time, and degree of damage to the silicon chip to help operators quickly locate and handle the problem.

[0098] Furthermore, if the operator determines that the current alarm is a false alarm, the false alarm information can be transmitted back to the logic processing unit through the feedback mechanism. The logic processing unit will record the false alarm information and optimize the alarm logic based on this information to reduce the probability of future false alarms. For example, by adjusting the sensitivity of the sensor or improving the parameters of the algorithm, the false alarm rate can be effectively reduced. It is understandable that this feedback mechanism enables the system to continuously learn and adapt, improving its own alarm accuracy and intelligence level.

[0099] On the contrary, if the operator confirms that there is a missing or damaged silicon wafer, the system will start the abnormal handling process. This process may include automatically stopping the transmission line, notifying maintenance personnel for maintenance, and recording abnormal events to ensure the safety and stability of the production process. It should be understood that the design of these steps should take into account the convenience of actual operation and the safety of the system to ensure that production can be resumed in the shortest time.

[0100] The benefit of this embodiment is that, by introducing a detailed interactive interface and feedback mechanism, the interaction between the operator and the system is enhanced, the accuracy and efficiency of alarm processing are improved, the interference of false alarms on production is reduced, and the operational reliability of the equipment is improved.

[0101] Embodiment 7

[0102] In order to solve the problem of silicon wafer transmission after the operator confirms the alarm, this embodiment further optimizes the silicon wafer transmission operation method after the confirmation instruction. Specifically, when the operator confirms a certain alarm signal through the control unit, the system will suspend the current silicon wafer transmission process to avoid further losses caused by abnormal conditions. This suspension mechanism is designed to temporarily stop the transmission operation of the relevant storage area, rather than stopping the entire equipment, thereby reducing the impact on the normal production process.

[0103] Furthermore, the control unit will adjust the transmission path of subsequent silicon wafers according to the abnormal conditions confirmed by the operator. This method includes calculating the optimal transmission order of the remaining silicon wafers, and giving priority to transmitting silicon wafers with longer distances and lower risks. It should be understood that through this path adjustment, equipment wear caused by repeated operations in the same area can be reduced, and it is ensured that the silicon wafers avoid passing through known abnormal areas again during transmission as much as possible. The method for calculating the optimal transmission order can be based on the shortest path algorithm in graph theory, such as the Dijkstra algorithm or the Floyd-Warshall algorithm, to ensure the optimization effect of the path.

[0104] After adjusting the transmission path, the system will control the movement of the X / Y / Z transfer module to transfer the silicon wafers from the large belt flow channel back to the flower basket or transfer new silicon wafers from the flower basket to the large belt flow channel until all abnormal situations are handled. The design of the X / Y / Z transfer module should have high precision and high reliability to ensure the safety of silicon wafers during transmission. For example, the transfer module can use high-speed servo motors and precision guide rails to fix the silicon wafers during transmission through vacuum adsorption to avoid slipping and damage.

[0105] Furthermore, the system will continuously update the silicon wafer status information of each storage area during the transmission process. The logic processing unit will dynamically adjust the alarm logic based on this updated status information. If a new abnormal situation is found during the subsequent transmission process, the system will immediately generate a new alarm signal and prompt the operator to confirm it through the control unit. It can be understood that this dynamic adjustment mechanism can ensure the timeliness and accuracy of alarm processing and avoid greater production losses due to delayed response.

[0106] The benefit of this embodiment is that by pausing the current transmission process, adjusting the transmission path and controlling the action of the X / Y / Z transplanting module, equipment wear and production interruptions caused by silicon wafer abnormalities are effectively reduced, thereby improving the operating efficiency of the system.

[0107] Embodiment 8

[0108] In order to solve the problem of false alarms caused by single sensor failure, this embodiment further optimizes the steps of the method for monitoring abnormal flow of silicon wafers in the track. Specifically, this embodiment introduces a data fusion method based on a time window to integrate detection data from different storage areas to reduce the occurrence of false alarms. This method comprehensively analyzes the alarm records within a fixed time window and uses a fuzzy logic algorithm to evaluate the credibility of the anomaly.

[0109] The specific implementation is as follows: the system is in each storage area P 1 , P 2 , P 3 …P nEach of the sensors is equipped with optical sensors, which detect the status of the silicon wafer in real time and generate alarm signals. When multiple sensors detect anomalies simultaneously or successively within a fixed time window, the system will use fuzzy logic algorithms to conduct a comprehensive evaluation of these alarm records. The calculation formula of the fuzzy logic algorithm is:

[0110]

[0111] Among them, n 1 ,n 2 ,…,n k is the number of alarms in different storage areas within the time window, w 1 ,w 2 ,…,w k is the corresponding weight, and T is the abnormality determination threshold. If C>T, it is considered that there is a high-confidence abnormality in the window.

[0112] Furthermore, this method can effectively reduce false alarms caused by single sensor failure. For example, if only one sensor detects an anomaly in a certain area, and other sensors do not detect an anomaly in the same time window, the fuzzy logic algorithm will reduce the credibility of the alarm. On the contrary, if multiple sensors detect the same anomaly in the same time window, the fuzzy logic algorithm will increase the credibility of the alarm.

[0113] In addition, this embodiment also considers the self-check function of the sensor. The logic processing unit will regularly perform self-checks on each optical sensor to ensure that it is in normal working condition. If a sensor is found to be faulty, the system will automatically adjust its weight or shield the alarm signal of the sensor to avoid false alarms caused by faulty sensors. It should be understood that this self-check function can further improve the stability of the system.

[0114] Furthermore, through the fusion results of all alarm data within the time window, the logic processing unit generates a processed alarm signal and transmits it to the control unit. The control unit prompts the operator to confirm through the interactive interface and performs subsequent silicon wafer transfer operations based on the operator's feedback. This not only improves the accuracy of alarm processing, but also reduces the number of operator interferences. It can be understood that the introduction of fuzzy logic algorithms can make the system more intelligent and flexible in processing alarm information, avoid false alarms caused by single sensor failures, and improve the reliability of alarm processing.

[0115] The benefit of this embodiment is that, through the time window-based data fusion method and fuzzy logic algorithm, the false alarms caused by single sensor failure are significantly reduced, the accuracy and reliability of alarm processing are improved, the risk of operator misoperation is reduced, and production efficiency is improved.

[0116] Embodiment 9

[0117] In order to solve the overall implementation problem of the silicon wafer flow abnormality monitoring system in the track, this embodiment discloses a silicon wafer flow abnormality monitoring system in the track. Specifically, the monitoring system includes a large belt flow channel and a small belt line. The large belt flow channel is provided with multiple storage areas P 1 , P 2 , P 3 …P n , used to store silicon wafers, each storage area is equipped with an optical sensor to detect the status of the silicon wafers in real time.

[0118] Furthermore, the system also includes an X / Y / Z transfer module for transferring the silicon wafers between the flower basket and the large belt flow channel. The transfer module uses vacuum adsorption to ensure that the silicon wafers will not slip during the transfer process, thereby improving the stability and safety of the transfer. The design of the vacuum adsorption system should have the characteristics of high efficiency and low power consumption. For example, a multi-stage vacuum pump and precision control valve can be used to ensure the stability of the adsorption force and the effective use of energy.

[0119] Furthermore, the system further includes a logic processing unit for receiving the silicon wafer status detection information and processing the alarm information through the monitoring method described in the above embodiment to avoid repeated alarms caused by the change of the silicon wafer position. The design of the logic processing unit should have the characteristics of high performance and low latency. For example, a multi-core processor and a real-time operating system can be used to ensure the timeliness and accuracy of data processing.

[0120] Furthermore, the system includes a control unit for receiving the processed alarm signal and prompting the operator to confirm. The control unit displays abnormal monitoring data and operation instructions in real time through an interactive interface to assist the operator in responding to and handling abnormal situations. The design of the interactive interface should be high-definition and user-friendly. For example, a touch screen and a graphical user interface (GUI) can be used to improve the operator's experience.

[0121] Furthermore, the system also has an exception handling process module, which starts the corresponding exception handling process according to the results confirmed by the operator. These processes include steps such as automatically recording exception information and notifying maintenance personnel to perform maintenance. The design of the exception handling process module should have the characteristics of automation and modularity. For example, event-driven architecture and modular design can be adopted to ensure the efficiency and flexibility of the processing process.

[0122] In addition, the system also has log recording and data analysis functions. The logic processing unit will record the detailed information and processing results of each alarm in a log file for subsequent analysis and optimization. The data analysis function can analyze historical alarm data based on machine learning algorithms to identify potential failure modes and optimization points, further improving the performance and reliability of the system.

[0123] It is understandable that through the optimized design of these components and functions, the monitoring system can more efficiently and reliably complete the detection and processing tasks of silicon wafer flow anomalies, and improve the stability and efficiency of production.

[0124] Through this embodiment, the monitoring system can not only realize real-time detection and alarm processing of silicon wafer status, but also has a high degree of automation and intelligence, which improves the work efficiency of operators, reduces production interruptions and equipment wear caused by abnormal situations, and improves the overall production level.

[0125] Embodiment 10

[0126] In order to solve the storage problem of the abnormal silicon wafer flow monitoring program in the track, this embodiment discloses a non-volatile storage medium. The storage medium stores the abnormal silicon wafer flow monitoring program in the track. When the monitoring program is executed by the processor, real-time monitoring and alarm processing of the silicon wafer status can be achieved.

[0127] Specifically, the readable storage medium may be a flash memory, a hard disk, an optical disk, or a network storage device. The selection of the storage medium should take into account factors such as storage capacity, read / write speed, and reliability to ensure efficient operation of the monitoring program. For example, a large-capacity SSD solid-state hard disk may be used to provide sufficient storage space and fast read / write speed.

[0128] Furthermore, the design of the monitoring program should be modular and highly scalable. The program can be divided into multiple functional modules, such as silicon chip status detection module, alarm processing module, path optimization module, data fusion module, etc. Each module is responsible for different functions and communicates and exchanges data through message queues or event-driven methods. Modular design not only facilitates program development and maintenance, but also enables flexible functional expansion according to actual needs.

[0129] Furthermore, the monitoring program should also have high reliability and fault tolerance. For example, the program can adopt a multi-threaded or multi-process design to ensure that the system can still operate normally when a single thread or process fails. In addition, the program should also have exception handling and logging functions. When a silicon chip abnormality is detected, an alarm signal can be generated in time, and detailed abnormal information can be recorded in a log file for subsequent analysis and processing.

[0130] Furthermore, the system should also have remote monitoring and management functions. Through network connection, operators and maintenance personnel can remotely access the monitoring program on the storage medium, view the current silicon wafer status and alarm information, and perform remote operation and management. This not only improves the convenience of operation, but also enables rapid response in emergency situations and reduces the time of production interruption. It is understandable that the design of remote monitoring and management functions should take into account network security and data privacy protection, and adopt encrypted communication and access control mechanisms to ensure the security of the system.

[0131] In addition, this embodiment also takes into account the self-update and self-learning capabilities of the program. When the system detects a new abnormal pattern or optimization point, the monitoring program can automatically adjust the algorithm parameters and logic through online learning or self-update mechanism to improve the accuracy and efficiency of alarm processing. The self-update mechanism can be based on a version control system, such as Git, to ensure that the program is always up to date by regularly checking for update packages or manually triggering updates.

[0132] The above embodiments further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for monitoring abnormal flow of silicon wafers in a track, which is applied to the alarm processing of missing or damaged silicon wafers in the flow channel, characterized in that: The monitoring method comprises: A large belt flow channel and a small belt line are provided, wherein the large belt flow channel is provided with a plurality of storage areas P1, P2, P3, ..., P n , used to store the silicon wafers, each of the storage areas is provided with an optical sensor for real-time detection of the existence and circulation of the silicon wafers; The silicon wafers are transferred between the flower basket and the large belt flow channel by an X / Y / Z transfer module to ensure that the silicon wafers can be accurately transferred from one process stage to another; During the silicon wafer transmission process, real-time monitoring of the status information of the silicon wafer in each storage area, including the location, access status and physical status; When it is detected that the silicon wafer in a certain storage area is missing or damaged, an initial alarm signal is generated, the initial alarm signal is transmitted to a logic processing unit, an alarm process is performed, and a subsequent transmission operation of the silicon wafer is adjusted; During the transmission process, the status information of the silicon wafers in each storage area is continuously updated, and the alarm logic is adjusted according to the updated status information.

2. The method for monitoring abnormal flow of silicon wafers in a track according to claim 1, characterized in that: The steps of the alarm processing include: Recording the storage area where the initial alarm signal occurs and the unique identification information of the silicon wafer, including the batch number and the process flow section number; Analyze the transmission path of the silicon wafer, and determine whether the current alarm is a repeated alarm by combining the built-in path algorithm with the historical alarm records. The path algorithm stores the flow direction of the silicon wafer in a tree structure, and each node represents a storage area. When an abnormal state of a node is detected, the path algorithm traces the state of the upstream and downstream nodes of the node to avoid repeated alarms due to position changes during the transmission of the silicon wafer; According to the analysis results, the logic processing unit generates a processing alarm signal and an operation flow; The processed alarm signal is transmitted to the control unit to prompt the operator to confirm and operate.

3. The method for monitoring abnormal flow of silicon wafers in a track according to claim 1, characterized in that: The step of the optical sensor detecting the state of the silicon wafer comprises: One or more optical sensors are arranged along the large belt flow path, each optical sensor group includes at least two photosensitive elements at different angles, and is used to obtain multi-angle image data of the silicon wafer; The acquired image data is preprocessed through the image processing module, including grayscale and filtering steps, to eliminate ambient light interference and improve detection accuracy; A deep learning algorithm is used to analyze the preprocessed image data to determine whether the storage area contains silicon wafers. If so, it is further determined whether the silicon wafers have signs of damage. The deep learning algorithm uses a convolutional neural network model and is trained using a large amount of pre-labeled silicon wafer image data to achieve high-precision silicon wafer status recognition.

4. The method for monitoring abnormal flow of silicon wafers in a track as claimed in claim 3, characterized in that: The deep learning algorithm also includes: The convolutional neural network model is pre-trained using a large amount of image data from the semiconductor processing technology, and then fine-tuned using a small amount of image data from the CVD process to accelerate the convergence speed of the model; A mechanism for dynamically adjusting the learning rate of the neural network is adopted to dynamically adjust the learning rate according to the current training error. The adjustment calculation formula of the learning rate is: Among them, α is the learning rate, α0 is the initial learning rate, α min is the minimum learning rate, T is the preset maximum number of training times, t is the current number of training times, and λ is the exponential decay speed parameter.

5. The method for monitoring abnormal flow of silicon wafers in a track according to claim 1, characterized in that: The logic processing unit predicts the state of the next storage area of ​​the silicon wafer by using a state prediction module and a Kalman filter before generating a processing alarm signal. The state prediction module predicts the future position state of the silicon wafer in real time based on the current position variable, speed variable and acceleration variable of the silicon wafer, combined with historical transmission data and model parameters; By comparing the predicted results with the actual detection results, it is determined whether it is a sensor false alarm or a silicon chip abnormality, and the alarm signal generation logic is adjusted according to the judgment result to reduce the occurrence of false alarms.

6. The method for monitoring abnormal flow of silicon wafers in a track according to claim 1, characterized in that: After prompting the operator to confirm, the monitoring method further comprises the following steps: The interactive interface displays alarm information and wafer status in real time, including the specific location, time, and degree of wafer damage, so that operators can quickly locate the problem. If the operator determines that the current alarm is a false alarm, the false alarm information is transmitted back to the logic processing unit through a feedback mechanism for learning and optimization to reduce the probability of future false alarms; If the operator confirms that the silicon wafer is missing or damaged, the exception handling process is initiated.

7. The method for monitoring abnormal flow of silicon wafers in a track according to claim 1, characterized in that: The specific steps of continuing to perform the silicon wafer transfer operation according to the operator's confirmation instruction include: Once the operator confirms the alarm, the current silicon wafer transfer process is suspended to avoid losses due to abnormalities; The control unit adjusts the transmission path of subsequent silicon wafers according to the confirmed abnormal situation, including calculating the optimal transmission order of the remaining silicon wafers and giving priority to silicon wafers with longer transmission distances and lower risks, thereby reducing equipment wear caused by repeated operations in the same area; By controlling the action of the X / Y / Z transplanting module, the transfer of the silicon wafers between the large belt flow channel and the flower basket is adjusted until all abnormal situations are handled.

8. The method for monitoring abnormal flow of silicon wafers in a track as claimed in claim 1, characterized in that: The monitoring method further comprises: By comprehensively analyzing the alarm records within a fixed time window, the fuzzy logic algorithm is used to evaluate the abnormal credibility. The calculation formula of the fuzzy logic algorithm is: If C>T, it is considered that there is a high-confidence abnormality in the window, where n1, n2, …, n k is the number of alarms in different storage areas within the time window, w1,w2,…,w k is the corresponding weight, T is the abnormality determination threshold; The fusion result of all the alarm data within the time window serves as the basis for the processing unit to generate an alarm signal.

9. A system for monitoring abnormal flow of silicon wafers in a track, characterized in that: The monitoring system comprises: A large belt flow channel and a small belt line, wherein the large belt flow channel is provided with a plurality of storage areas P1, P2, P3, ..., P n , used to store the silicon wafers, each of the storage areas is equipped with an optical sensor for real-time monitoring of the status of the silicon wafers; An X / Y / Z transfer module, used to transfer the silicon wafers between the flower basket and the large belt flow channel, wherein the transfer module uses a vacuum adsorption method to ensure that the silicon wafers do not slip during the transfer process; A logic processing unit, configured to receive wafer status detection information, and process the alarm information by the method for monitoring abnormal wafer flow in a track according to any one of claims 1 to 8, so as to avoid repeated alarms caused by position changes; A control unit is used to receive the processing alarm signal and prompt the operator to confirm, and perform subsequent silicon wafer flow operations according to the operator's feedback An interactive interface that displays abnormal monitoring data and operating instructions in real time, assisting operators in responding to and handling abnormal situations; An exception handling process module starts the corresponding exception handling process according to the result of operation confirmation, including automatically recording exception information and arranging maintenance operations.

10. A readable storage medium, characterized in that: The storage medium stores a program for monitoring abnormal flow of silicon wafers in a track. When the monitoring program is executed by the processor, the processor executes the steps of the method for monitoring abnormal flow of silicon wafers in a track as described in any one of claims 1 to 8.

Images