Semiconductor process anomaly processing method, device, equipment, medium and program product

Abstract

The application relates to a semiconductor process exception processing method, device, equipment, medium and program product. The method comprises the following steps: acquiring exception alarm information; performing feature extraction on the exception alarm information to obtain exception features; determining calling utility values corresponding to different service combinations based on the exception features, wherein each service combination comprises different services for performing exception diagnosis based on the exception alarm information; determining a target service combination based on the calling utility values, and performing exception diagnosis on the exception alarm information through each service in the target service combination to obtain an exception processing result. The method can improve the processing efficiency.

Description

Technical Field

[0001] This application relates to the field of semiconductor technology, and in particular to a method, apparatus, device, medium, and process product for handling semiconductor process anomalies. Background Technology

[0002] In the ultra-precise manufacturing process of semiconductor wafers, each wafer undergoes hundreds or even thousands of process steps, and its final yield highly depends on the real-time capture and accurate tracing of various anomalies. To this end, wafer fabs typically deploy yield and anomaly management systems, including yield analysis services, AI-based defect scanning services, wafer electrical parameter testing services, root cause analysis services, and standard baseline process path services for comparison. These systems each provide specialized data analysis capabilities for different dimensions of process anomalies, forming the technological foundation for ensuring production stability.

[0003] However, in current anomaly response practices, these specialized services lack effective automated collaboration mechanisms, forcing process integration engineers to rely on personal experience for "manual coordination." When a batch triggers an alert due to yield anomalies, the PIE (Process Integration Engineer) must log into four to five different independent systems sequentially, copy the batch numbers of the anomalies one by one, and check the yield distribution in the yield analysis service, identify defect morphologies in the AI ​​defect scanning service, verify electrical offsets in the wafer electrical parameter testing service, and then compare them with normal batches through the root cause analysis service and the standard baseline process path service. The entire process heavily relies on the PIE's memory of historical cases and on-site judgment.

[0004] This operational mode, which relies heavily on manual connections, results in a time-consuming and extremely inefficient analysis process. Summary of the Invention

[0005] Therefore, it is necessary to provide a semiconductor process anomaly handling method, apparatus, equipment, medium, and program product that can improve analysis efficiency in response to the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for handling semiconductor process anomalies, the method comprising:

[0007] Obtain abnormal alarm information;

[0008] The abnormal alarm information is subjected to feature extraction to obtain abnormal features;

[0009] Based on the abnormal characteristics, the invocation utility value corresponding to different service combinations is determined, and each service combination includes different services for abnormal diagnosis based on the abnormal alarm information.

[0010] The target service combination is determined based on the invocation utility value, and the abnormal alarm information is diagnosed by each service in the target service combination to obtain the abnormal handling result.

[0011] In one embodiment, determining the invocation utility value corresponding to different service combinations based on the anomaly characteristics includes:

[0012] Based on the aforementioned anomaly characteristics, the utility weight corresponding to each of the services is determined;

[0013] Based on at least one of the available data required for the service, similar historical cases corresponding to the abnormal features, and pre-detection results corresponding to the abnormal features, determine the information gain for each of the services.

[0014] Based on the utility weight and information gain of each service, the invocation utility value corresponding to different service combinations is determined.

[0015] In one embodiment, determining the utility weight corresponding to each service based on the anomaly characteristics includes:

[0016] Obtain the weighted rules from the expert rule base, wherein the weighted rules include service identifiers, judgment conditions, and target weights;

[0017] If the abnormal feature satisfies the judgment condition in the weighting rule, the utility weight of the service corresponding to the judgment condition is determined as the target weight.

[0018] In one embodiment, determining the utility weight corresponding to each service based on the anomaly characteristics includes:

[0019] Obtain a weight determination model based on statistical learning of historical cases from the sample;

[0020] The abnormal features are processed by the weight determination model to obtain the utility weights corresponding to each service.

[0021] In one embodiment, determining the information gain for each service based on at least one of the available data volume required by the service, similar historical cases corresponding to the anomaly feature, and pre-detection results corresponding to the anomaly feature includes:

[0022] The amount of data required by each of the services in the abnormal features is determined as the actual available data amount, and a data availability index value is obtained based on the actual available data amount and the amount of data required for the normal operation of the service.

[0023] Based on the abnormal features, a first number of similar historical cases are determined, and a second number of similar historical cases successfully locate the root cause using the target service is counted. A third number of similar historical cases use the target service, and a historical experience index value is obtained based on the second number and the third number.

[0024] The abnormal features are compared with normal process data to obtain an abnormal deviation index value used to characterize the pre-inspection result corresponding to the abnormal features.

[0025] The information gain corresponding to each service is determined based on at least one of the data availability index, historical experience index, and abnormal offset index.

[0026] In one embodiment, the method further includes:

[0027] When the service is started, the service information of the service is registered;

[0028] Before determining the invocation utility value corresponding to different service combinations based on the abnormal characteristics, the method further includes:

[0029] Obtain effective services based on service information provided by the service.

[0030] Different service combinations are generated based on effective services.

[0031] Secondly, this application also provides a semiconductor process anomaly handling apparatus, the apparatus comprising:

[0032] An abnormal alarm information acquisition module is used to acquire abnormal alarm information;

[0033] The feature extraction module is used to extract abnormal features from the abnormal alarm information;

[0034] The utility value determination module is used to determine the invocation utility value corresponding to different service combinations based on the abnormal characteristics, and each service combination includes different services for abnormal diagnosis based on the abnormal alarm information.

[0035] An exception handling module is used to determine a target service combination based on the call utility value, and to perform exception diagnosis on the exception alarm information through each service in the target service combination to obtain the exception handling result.

[0036] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method in any of the above embodiments.

[0037] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the methods in any of the above embodiments.

[0038] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0039] The aforementioned semiconductor process anomaly handling method, apparatus, equipment, medium, and program products acquire anomaly alarm information; extract features from the anomaly alarm information to obtain anomaly features; determine the invocation utility value corresponding to different service combinations based on the anomaly features, each service combination including different services for anomaly diagnosis based on the anomaly alarm information; determine a target service combination based on the invocation utility value, and perform anomaly diagnosis on the anomaly alarm information through each service in the target service combination to obtain anomaly handling results. This process, by determining the invocation utility value corresponding to different service combinations based on anomaly features, then determining the target service combination based on the invocation utility value, and performing anomaly diagnosis through the target service combination, requires no manual intervention throughout, thus improving diagnostic efficiency. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is an application environment diagram of a semiconductor process anomaly handling method in one embodiment;

[0042] Figure 2 This is a flowchart illustrating a semiconductor process anomaly handling method in one embodiment;

[0043] Figure 3 This is a flowchart of the steps for generating a utility value in one embodiment;

[0044] Figure 4 Here is a flowchart of the information gain determination step in one embodiment;

[0045] Figure 5 This is a structural block diagram of a semiconductor process anomaly handling device in one embodiment;

[0046] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0049] The semiconductor process anomaly handling method provided in this application embodiment can be applied to, for example... Figure 1 The server shown has an event monitoring layer coupled to the wafer fab's metrology equipment and machine sensors. This layer can receive wafer anomaly events in real time and generate alarm information. This alarm information can be uploaded to the task scheduling layer, which communicates with the service pool. The task scheduling layer acts as a central scheduling node, managing the service pool and scheduling services within it. Each service has independent reasoning capabilities and can exchange tasks with the task scheduling layer. The task scheduling layer and the service pool can communicate via the A2A protocol.

[0050] The event monitoring layer receives wafer anomaly events in real time and generates anomaly alarm information. The task scheduling layer extracts features from the anomaly alarm information to obtain anomaly characteristics. Based on the anomaly characteristics, it determines the invocation utility value corresponding to different service combinations. Each service combination includes different services used for anomaly diagnosis based on the anomaly alarm information. Based on the invocation utility value, it determines the target service combination. Finally, each service in the target service combination performs anomaly diagnosis on the anomaly alarm information to obtain the anomaly handling result, and sends the anomaly handling result back to the task scheduling layer. In this way, the entire anomaly diagnosis process does not require manual intervention, improving efficiency.

[0051] In one exemplary embodiment, such as Figure 2 As shown, a semiconductor process anomaly handling method is provided, which is applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 208. Wherein:

[0052] S202: Obtain abnormal alarm information.

[0053] The abnormal alarm information can be obtained through the event monitoring layer. Specifically, when abnormal events occur in the wafer fab's measurement equipment and machine sensors, the event monitoring layer can receive these abnormal events, generate corresponding abnormal alarm information, and send the abnormal alarm information to the task scheduling layer.

[0054] Specifically, measurement equipment and machine sensors can monitor monitored objects in the wafer fab, including at least one of key measurement parameters and machine status signals. The key measurement parameters include at least one of film thickness, critical dimension (CD), overlay error, and defect density. The machine status signals include at least one of fault detection and classification (FDC) indicators such as cavity pressure, RF power, temperature drift, and gas flow rate.

[0055] When the measuring equipment and machine sensor determine that at least one of the monitored objects exceeds the preset control limit or violates the statistical process control (SPC) rules (such as Western Electric Rules), it is determined to be an out-of-control (OOC) event.

[0056] Optionally, after the event monitoring layer obtains an abnormal event, it outputs standardized abnormal alarm information, which includes object identifier (e.g., wafer identifier or batch identifier), monitored object name and value, abnormal event timestamp, and at least one of the process steps to which the abnormality belongs.

[0057] Optionally, the event monitoring layer can send abnormal alarm information to the central scheduling node (Master Agent) of the task scheduling layer asynchronously, and the abnormal alarm information serves as the initial trigger event for subsequent collaborative analysis.

[0058] It should be noted that the event monitoring layer only serves as a bridge for anomaly detection, standardization, and the dissemination of anomaly alarm information to ensure seamless integration with existing wafer fab systems.

[0059] S204: Extract abnormal features from abnormal alarm information.

[0060] The abnormal features include at least one of the following: abnormality type, severity, spatial distribution, temporal characteristics, and parameters. In this embodiment, a pre-trained large model can be used to extract features from the abnormal alarm information to obtain the abnormal features. In other embodiments, abnormal features can also be obtained by extracting features from the abnormal alarm information based on pre-defined feature extraction rules.

[0061] S206: Determine the invocation utility value corresponding to different service combinations based on abnormal characteristics. Each service combination includes different services used for abnormal diagnosis based on abnormal alarm information.

[0062] The service is used for anomaly diagnosis based on abnormal alarm information. This service may include multiple core services such as integrated yield prediction analysis (YPA), intelligent defect classification (AI-Defect), golden path recommendation (GoldenPath), wafer resume analysis (WRA), and root cause analysis (RCA). In other embodiments, the service may also include others. This is only for illustrative purposes. These services are all for processing abnormal alarm information to obtain anomaly diagnosis results.

[0063] In this application, each service can be obtained, and then the service combination corresponding to each service can be obtained. Assuming that the services include service A, service B, and service C, the corresponding service combination includes a combination of service A, service B, or service C that includes only one service; it can also be a combination of two services, such as a combination of service A and service B, a combination of service A and service C, or a combination of service C and service B; it can also be a combination of three services, or a combination of service A, service B, and service C.

[0064] The task scheduling layer can calculate the invocation utility value corresponding to each service combination. For example, based on the anomaly characteristics, it can first calculate the information gain corresponding to each service separately, and then generate the invocation utility value of the service combination based on the services included in each service combination, the information gain corresponding to the service, and the weight corresponding to the service.

[0065] S208: Determine the target service composition based on the call utility value, and obtain the abnormal handling result by performing abnormal alarm information through each service in the target service composition.

[0066] In this embodiment, the task scheduling layer can determine the target service combination based on the call utility value. For example, it can only start the high-efficiency service combination based on the call utility value and skip the inefficient or informationless service combination. Optionally, in this application, the target service combination with the largest call utility value can be selected, and then the abnormal alarm information can be diagnosed through each service in the target service combination to obtain the abnormal handling result.

[0067] For ease of understanding, the above-mentioned Yield Prediction Analysis (YPA) is used to assess the degree of damage (SeverityScore) of abnormal events to the final yield based on historical yield models and current abnormal characteristics, and outputs the predicted yield loss value σ∈[0,1] and the key layers affected.

[0068] The Intelligent Defect Classification (AI-Defect) service receives defect images or ADC (Automatic Defect Classification) results and uses CV (Computer Vision) or VLM (Vision-Language Model) models to identify failure modes, such as particle, bridge, and open circuit. This service outputs defect types, spatial distribution maps, and a Top-K list of similar historical cases.

[0069] Wafer Resume Analysis is used to trace the complete process flow of all wafers in an abnormal batch and extract the Common Path Machine Set. This service outputs a machine list P={m1m2,...,mL} along with timestamps and recipe IDs for each process step.

[0070] The GoldenPath recommendation service compares the actual path with a predefined GoldenPath and calculates the Path Deviation Index (PDI). This service outputs the PDI score, deviation steps, and recommended regression path.

[0071] Root Cause Analysis (RCA) integrates device logs, parameter trends, and defect characteristics, using causal inference or association rule mining to generate a ranking of suspicious factors. This service outputs a list of candidate root causes {ci} and statistical scores for each candidate root cause.

[0072] Therefore, the diagnostic results can include at least one of the following: a credible root cause ranking (verified by physical paths), yield impact prediction, golden path deviation suggestions, defect pattern comparison, and Design of Experiments (DOE) parameter mapping. Each service sends the diagnostic results back to the task scheduling layer, which integrates the results to generate a diagnostic report.

[0073] The aforementioned semiconductor process anomaly handling method involves: acquiring anomaly alarm information; extracting features from the anomaly alarm information to obtain anomaly features; determining the call utility value corresponding to different service combinations based on the anomaly features, with each service combination including different services used for anomaly diagnosis based on the anomaly alarm information; determining the target service combination based on the call utility value, and then using each service in the target service combination to perform anomaly diagnosis on the anomaly alarm information to obtain the anomaly handling result. This method determines the call utility value corresponding to different service combinations based on anomaly features, then determines the target service combination based on the call utility value, and performs anomaly diagnosis using the target service combination. The entire process requires no manual intervention, improving diagnostic efficiency.

[0074] In some alternative embodiments, combined with Figure 3 As shown, Figure 3 This is a flowchart of a step in generating a call utility value in one embodiment. This step, which determines the call utility value corresponding to different service combinations based on anomaly characteristics, includes:

[0075] S302: Determine the utility weight of each service based on anomaly characteristics.

[0076] The utility weights for each service can be obtained based on expert rule bases or statistical learning of historical sample cases. No specific limitations are made here. Under different abnormal characteristics, the utility weights for each service are different.

[0077] S304: Determine the information gain for each service based on at least one of the following: the amount of available data required for the service, similar historical cases corresponding to the anomaly features, and the pre-detection results corresponding to the anomaly features.

[0078] The information gain corresponding to the service can be generated based on at least one of the following: the amount of available data required for the service, similar historical cases corresponding to the anomaly characteristics, and the pre-detection results corresponding to the anomaly characteristics. The amount of available data required for the service can be used to generate a data availability index value; similar historical cases corresponding to the anomaly characteristics can be used to generate a historical experience index value; and the pre-detection results corresponding to the anomaly characteristics can be used to generate an anomaly offset index value. The data availability index value is used to determine whether the data in the anomaly alarm information is available; the higher the data availability, the higher the score of this index value. The historical experience index value is used to determine whether historical experience supports the data in the anomaly alarm information; if similar anomalies are found, the historical experience index value is determined based on whether the root cause was successfully located in the past. The anomaly offset index value is used for lightweight pre-detection results to quickly analyze whether a significant offset has occurred.

[0079] Finally, the information gain for each service is determined based on at least one of the data availability index, historical experience index, and abnormal offset index.

[0080] S306: Determine the invocation utility value for different service combinations based on the utility weight and information gain of each service.

[0081] After obtaining the utility weight and information gain of each service, the invocation utility value of the service combination can be determined based on the services in the service combination.

[0082]

[0083] in, To invoke utility value, For utility weight, For information gain, s represents service. Let S be the service composition, and arg max be the target service composition for which the maximum invocation utility value is sought.

[0084] In some optional embodiments, the utility weight of each service is determined based on the anomaly characteristics, including: obtaining each weight rule in the expert rule base, the weight rule including the service identifier, the judgment condition and the target weight; and determining the utility weight of the service with the service identifier corresponding to the judgment condition as the target weight when the anomaly characteristics meet the judgment condition in the weight rule.

[0085] The weight rules in this expert rule base can be pre-defined, and if there are conflicts between weight rules, the available weight rules can be determined based on the priority of the weight rules.

[0086] The weighting rule includes a service identifier, a judgment condition, and a target weight. The service identifier is the service to which the weighting rule applies. The judgment condition is used to determine whether the value corresponding to the abnormal feature meets the judgment condition, thereby determining the utility weight of the service. That is, if the abnormal feature meets the judgment condition in the weighting rule, the utility weight of the service with the service identifier corresponding to the judgment condition is determined as the target weight.

[0087] Table 1 shows the abnormal feature values ​​corresponding to the abnormal features in one embodiment. In other embodiments, the abnormal feature values ​​corresponding to the abnormal features may be different. This is only for illustrative purposes.

[0088] Table 1

[0089]

[0090] In some optional embodiments, referring to Table 2, which shows the distance of the weighting rule in one embodiment, the weighting rule can be different in other embodiments; this is only for illustrative purposes.

[0091] Table 2

[0092]

[0093] In other alternative embodiments, the utility weights corresponding to each service are determined based on the abnormal features, including: obtaining a weight determination model based on statistical learning of historical cases of samples; and processing the abnormal features through the weight determination model to obtain the utility weights corresponding to each service.

[0094] In this application, a weight determination model can also be derived through statistical learning of historical cases. After obtaining abnormal features, these features are then input into the weight determination model to obtain the utility weights of each service output by the weight determination model.

[0095] The weight determination model is generated based on the anomaly features corresponding to historical cases, the weights corresponding to these anomaly features, and the actual diagnostic results. The model's training samples include the anomaly features of each service at the time of the failure, as well as the diagnostic results annotated by operations and maintenance experts. The model is trained using either learning by ranking or linear weighted regression. During training, the model continuously attempts to assign a weight coefficient to each anomaly feature. For example, the utility weights corresponding to different anomaly features are trained by analyzing the utility weights corresponding to the anomaly features of similar historical cases to obtain the utility weights for different historical cases.

[0096] The determination of the utility weights described above is merely an example; other methods may be used in other embodiments.

[0097] In some alternative embodiments, combined with Figure 4 As shown, Figure 4 This is a flowchart of an information gain determination step in one embodiment. This information gain step, based on at least one of the following: the amount of available data required by the service, similar historical cases corresponding to the anomaly features, and pre-detection results corresponding to the anomaly features, determines the information gain for each service, including:

[0098] S402: Determine the amount of data required by each service in the abnormal characteristics as the actual available data amount, and obtain the data availability index value based on the actual available data amount and the amount of data required for the normal operation of the service.

[0099] The data availability metric indicates whether the data required for the service exists; the higher the data availability, the higher the score of the data availability metric.

[0100] The data availability metric value can be calculated using the following formula:

[0101]

[0102] The actual usable data volume refers to the percentage of complete and readable data fields required by the service during the current abnormal event. The theoretical maximum data volume is the total number of data fields required for the service to operate normally.

[0103] S404: Determine the first number of similar historical cases based on abnormal features, and count the second number of similar historical cases that successfully located the root cause using the target service, and count the third number of similar historical cases that used the target service. Based on the second and third numbers, obtain the historical experience index value.

[0104] The historical experience index value is used to determine whether historical experience supports the conclusion that the root cause of similar anomalies has been successfully located in the past.

[0105] The historical experience indicator value is calculated as follows:

[0106]

[0107] The first quantity is Top-K, the second quantity is the number of similar historical cases that successfully located the root cause using the target service, which is the numerator of the above formula, and the third quantity is the number of similar historical cases that used the target service, which is the denominator of the above formula.

[0108] S406: Compare the abnormal features with normal process data to obtain the abnormal offset index value used to characterize the pre-inspection result corresponding to the abnormal features.

[0109] The abnormal offset index value is a lightweight pre-detection result used to quickly analyze whether significant offsets are found. Specifically, it quickly compares the key parameters of the current abnormal batch with those of the normal batch (Golden Batch) in the normal process data, and calculates statistical indicators such as Z-Score and offset as abnormal offset index values.

[0110] S408: Determine the information gain for each service based on at least one of the data availability index, historical experience index, and abnormal offset index.

[0111] Information gain can be determined based on at least one of the following: data availability metric, historical experience metric, and outlier metric. For example:

[0112]

[0113] in, For information gain, , as well as The weights for data availability metrics, historical experience metrics, and outlier metrics can be equal or determined based on expert experience. This is a data availability metric value. These are historical experience indicators. This is the value of the abnormal offset indicator.

[0114] In some optional embodiments, the method further includes: registering service information of the service when the service starts; before determining the invocation utility value corresponding to different service combinations based on the anomaly characteristics, the method further includes: obtaining valid services based on the service information of the service; and generating different service combinations based on the valid services.

[0115] In this application, each service registers its service information with the central scheduling node upon startup. This service information can be a structured metadata object, including the service name, etc., which is not limited here. The central scheduling node can query the service information at runtime and dynamically match available service combinations according to event requirements. The timing of service startup can be arbitrary, but to ensure the normal operation of the system, at least one service can be ensured to start before determining the invocation utility value corresponding to different service combinations based on anomaly characteristics. The startup time of the service is not required. In practical applications, generally all services are started before determining the invocation utility value corresponding to different service combinations based on anomaly characteristics.

[0116] Before determining the invocation utility value corresponding to different service combinations based on anomaly characteristics, valid services can be obtained based on the service information. Therefore, it is only necessary to obtain the service combinations corresponding to valid services, without considering invalid services, so as to reduce the amount of computation. Finally, only the invocation utility of generating different service combinations based on valid services needs to be calculated.

[0117] In this application, the Google A2A (Agent-to-Agent) communication protocol is adopted, and efficient serialization is achieved based on remote procedure call (gRPC) and protocol buffers.

[0118] The communication process includes: after receiving semiconductor anomaly alarm information (SNS), the central scheduling node (Master Agent) parses the request and queries the service information for target services that match its capabilities; after dynamically filtering target services, it initiates a call via the A2A protocol; after the called target service completes its analysis, the target service sends the result back to the central scheduling node (Master Agent); the central scheduling node (Master Agent) listens to all result topics and aggregates and outputs them according to the ID of the anomaly alarm information. Optionally, this embodiment supports timeout retries, load balancing, and version compatibility management.

[0119] In the above embodiments, based on real-time anomaly characteristics, each service is dynamically scored using a utility function. Only services with utility values ​​exceeding a threshold are invoked, achieving true on-demand analysis. Computational resources are precisely allocated to high-value analytical tasks. Server load, energy consumption, and cloud service costs are significantly reduced. Diagnostic time is determined by a few high-efficiency services, resulting in a significantly shorter and more predictable cycle time. Automated processes eliminate human waiting. Expert rules and historical experience are embedded in the utility function, becoming maintainable and iterative digital assets. The system possesses the ability to continuously learn and self-optimize.

[0120] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0121] Based on the same inventive concept, this application also provides a semiconductor process anomaly handling apparatus for implementing the semiconductor process anomaly handling method described above. The solution provided by this apparatus is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the semiconductor process anomaly handling apparatus provided below can be found in the limitations of the semiconductor process anomaly handling method described above, and will not be repeated here.

[0122] In one exemplary embodiment, such as Figure 5 As shown, a semiconductor process anomaly handling device is provided, comprising: an anomaly alarm information acquisition module 501, a feature extraction module 502, a utility value determination module 503, and an anomaly handling module 504, wherein:

[0123] The abnormal alarm information acquisition module 501 is used to acquire abnormal alarm information;

[0124] The feature extraction module 502 is used to extract abnormal features from abnormal alarm information.

[0125] The utility value determination module 503 is used to determine the call utility value corresponding to different service combinations based on the abnormal characteristics. Each service combination includes different services used for abnormal diagnosis based on abnormal alarm information.

[0126] The exception handling module 504 is used to determine the target service combination based on the call utility value, and to obtain the exception handling result by performing exception diagnosis on the exception alarm information through each service in the target service combination.

[0127] In some optional embodiments, the utility value determination module 503 is specifically used to determine the utility weight of each service based on the anomaly features; determine the information gain of each service based on at least one of the available data required by the service, similar historical cases corresponding to the anomaly features, and pre-detection results corresponding to the anomaly features; and determine the call utility value corresponding to different service combinations based on the utility weight of each service and the information gain.

[0128] In some optional embodiments, the utility value determination module 503 is specifically used to obtain each weight rule in the expert rule base. The weight rule includes a service identifier, a judgment condition, and a target weight. When the abnormal feature meets the judgment condition in the weight rule, the utility weight of the service identifier corresponding to the judgment condition is determined as the target weight.

[0129] In some optional embodiments, the utility value determination module 503 is specifically used to obtain a weight determination model based on statistical learning of historical cases of samples; and to process abnormal features through the weight determination model to obtain the utility weights corresponding to each service.

[0130] In some optional embodiments, the utility value determination module 503 is specifically used to determine the amount of data required by each service in the abnormal features as the actual available data amount, and obtain a data availability index value based on the actual available data amount and the amount of data required for normal service operation; determine a first number of similar historical cases based on the abnormal features, and count a second number of similar historical cases that successfully located the root cause using the target service, and count a third number of similar historical cases that used the target service, and obtain a historical experience index value based on the second and third numbers; compare the abnormal features with normal process data to obtain an abnormal offset index value used to characterize the pre-detection result corresponding to the abnormal features; and determine the information gain corresponding to each service based on at least one of the data availability index value, historical experience index value, and abnormal offset index value.

[0131] In some optional embodiments, the above apparatus further includes a registration module for registering service information of the service when the service is started.

[0132] The service composition generation module is used to obtain valid services based on service information and generate different service compositions based on valid services.

[0133] Each module in the aforementioned semiconductor process anomaly handling device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0134] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores various data involved in the aforementioned methods. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a semiconductor process exception handling method.

[0135] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0136] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0137] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0138] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0139] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0140] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0141] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0142] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for handling semiconductor process anomalies, characterized in that, The method includes: Obtain abnormal alarm information; The abnormal alarm information is subjected to feature extraction to obtain abnormal features; Based on the abnormal characteristics, the invocation utility value corresponding to different service combinations is determined, and each service combination includes different services for abnormal diagnosis based on the abnormal alarm information. Based on the invoke utility value, a target service combination is determined, and anomaly diagnosis is performed on the abnormal alarm information through each service in the target service combination to obtain the anomaly handling result. The step of determining the invocation utility value corresponding to different service combinations based on the abnormal characteristics includes: Based on the aforementioned anomaly characteristics, the utility weight corresponding to each of the services is determined; Based on at least one of the available data volume required for the service, similar historical cases corresponding to the anomaly feature, and pre-detection results corresponding to the anomaly feature, the information gain corresponding to each service is determined, including: determining the data volume required for each service in the anomaly feature as the actual available data volume; obtaining a data availability index value based on the actual available data volume and the data volume required for the normal operation of the service; determining a first number of similar historical cases based on the anomaly feature, and counting a second number of similar historical cases that successfully located the root cause using the target service, and counting a third number of similar historical cases that used the target service; obtaining a historical experience index value based on the second number and the third number; comparing the anomaly feature with normal process data to obtain an anomaly offset index value used to characterize the pre-detection results corresponding to the anomaly feature; and determining the information gain corresponding to each service based on at least one of the data availability index value, historical experience index value, and anomaly offset index value. Based on the utility weight and information gain of each service, the invocation utility value corresponding to different service combinations is determined.

2. The method according to claim 1, characterized in that, The determination of the utility weight corresponding to each service based on the anomaly characteristics includes: Obtain the weighted rules from the expert rule base, wherein the weighted rules include service identifiers, judgment conditions, and target weights; If the abnormal feature satisfies the judgment condition in the weighting rule, the utility weight of the service corresponding to the judgment condition is determined as the target weight.

3. The method according to claim 1, characterized in that, The determination of the utility weight corresponding to each service based on the anomaly characteristics includes: Obtain a weight determination model based on statistical learning of historical cases from the sample; The abnormal features are processed by the weight determination model to obtain the utility weights corresponding to each service.

4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: When the service is started, the service information of the service is registered; Before determining the invocation utility value corresponding to different service combinations based on the abnormal characteristics, the method further includes: Obtain effective services based on service information provided by the service. Different service combinations are generated based on effective services.

5. A semiconductor process anomaly handling device, characterized in that, The device includes: An abnormal alarm information acquisition module is used to acquire abnormal alarm information; The feature extraction module is used to extract abnormal features from the abnormal alarm information; The utility value determination module is used to determine the invocation utility value corresponding to different service combinations based on the abnormal characteristics, and each service combination includes different services for abnormal diagnosis based on the abnormal alarm information. An exception handling module is used to determine a target service combination based on the call utility value, and to perform exception diagnosis on the exception alarm information through each service in the target service combination to obtain the exception handling result; The utility value determination module is further configured to: determine the utility weight corresponding to each service based on the anomaly feature; determine the information gain corresponding to each service based on at least one of the available data volume required by the service, similar historical cases corresponding to the anomaly feature, and pre-detection results corresponding to the anomaly feature, including: determining the data volume required by each service in the anomaly feature as the actual available data volume, obtaining a data availability index value based on the actual available data volume and the data volume required for normal operation of the service; determining a first number of similar historical cases based on the anomaly feature, and counting a second number of similar historical cases that successfully located the root cause using the target service, counting a third number of similar historical cases that used the target service, obtaining a historical experience index value based on the second number and the third number; comparing the anomaly feature with normal process data to obtain an anomaly offset index value used to characterize the pre-detection results corresponding to the anomaly feature; determining the information gain corresponding to each service based on at least one of the data availability index value, historical experience index value, and anomaly offset index value; and determining the call utility value corresponding to different service combinations based on the utility weight and information gain corresponding to each service.

6. The apparatus according to claim 5, characterized in that, The utility value determination module is specifically used for: Obtain the weighted rules from the expert rule base, wherein the weighted rules include service identifiers, judgment conditions, and target weights; If the abnormal feature satisfies the judgment condition in the weighting rule, the utility weight of the service corresponding to the judgment condition is determined as the target weight.

7. The apparatus according to claim 5, characterized in that, The utility value determination module is specifically used for: Obtain a weight determination model based on statistical learning of historical cases from the sample; The abnormal features are processed by the weight determination model to obtain the utility weights corresponding to each service.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images