Fuzzy knowledge graph construction method related to wafer manufacturing process, medium, equipment and product

Abstract

The invention provides a fuzzy knowledge graph construction method related to a wafer manufacturing process, and relates to the technical field of fuzzy knowledge graphs, and the method comprises the steps: employing a three-dimensional convolutional network to extract a spatial defect feature X of a scanning electron microscope image, carrying out the coding of sensor time series data through a bidirectional LSTM, and obtaining a feature set Y, obtaining electrical performance parameters of the wafer based on impedance spectroscopy analysis from the text to obtain a feature set Z; building fuzzy knowledge based on the text to reinforce the feature sets X, Y and Z, building a fuzzy feature set and a time sequence feature set, building a weighted scoring model, and screening a candidate set from the text according to the weighted scoring model; information is extracted from the candidate set, and a tetrad structure G = (entities, relationships, time attributes and fuzzy rules) is constructed; and taking G as input, generating a time sequence fuzzy rule through a recursive reconstruction contribution method, and incrementally updating the knowledge graph. According to the method, the characterization bottleneck of space-time coupling and uncertainty correlation in a traditional method is broken through.

Description

Technical Field

[0001] This invention relates to the field of fuzzy knowledge graph technology, and in particular to a method, medium, equipment, and product for constructing a fuzzy knowledge graph related to wafer manufacturing processes. Background Technology

[0002] Complex nonlinear coupling exists between wafer manufacturing defects and process parameters. Current research has expanded from single-sensor data (such as temperature and pressure) to multimodal data fusion, including SEM images, spectral data, and equipment logs. Existing methods (such as feature stitching and shared encoders) only achieve shallow statistical alignment between modalities and cannot capture deep spatiotemporal coupling mechanisms. The implicit correlation between spatial distribution patterns of defects in SEM images (such as ring clusters) and temporal fluctuations in chamber pressure (high-frequency jitter) is ignored. Existing Bayesian networks rely on manually set prior probabilities and cannot automatically quantify the fuzzy mapping of "process fluctuations → defect probability". Engineers' fuzzy experience (such as "plasma instability easily leads to edge residue") lacks mathematical expression and cannot guide model optimization. In summary, current methods face a double limitation: First, a single data modality is insufficient to capture the multidimensional correlation of defect-sensitive parameters. For example, there may be an implicit correlation between spatial defect patterns in SEM images and temporal drift of sensors; second, deterministic modeling paradigms cannot quantify the fuzzy impact of process fluctuations on defect generation (such as the probabilistic causal relationship in "a decrease in etching rate may lead to a short circuit"). Although existing research attempts to integrate multi-source data, the modeling of cross-modal spatiotemporal correlation features remains at the shallow alignment stage, and the lack of a feature optimization mechanism guided by fuzzy knowledge limits the sensitivity representation ability of the defect-process mapping model. Summary of the Invention

[0003] The purpose of this invention is to address the problem of nonlinear coupling of multi-source data in wafer manufacturing processes by proposing a method for constructing a fuzzy knowledge graph related to wafer manufacturing processes, comprising the following steps: S1. Acquire scanning electron microscope images, sensor timing data, parameter settings, and text of the rule base during the wafer manufacturing process; S2. A three-dimensional convolutional network is used to extract spatial defect features from scanning electron microscope images to obtain feature set X. The sensor time series data is encoded by a bidirectional LSTM to obtain feature set Y. The electrical performance parameters of the wafer based on impedance spectrum analysis are obtained from the text to obtain feature set Z. S3. Strengthen feature sets X, Y, and Z by constructing fuzzy knowledge based on text; S4. Construct a fuzzy feature set based on the enhanced feature sets X, Y and Z, extract a temporal feature set from the enhanced feature set Y, construct a weighted scoring model based on the fuzzy feature set and the temporal feature set, and filter candidate sets from the text according to the weighted scoring model; S5. Extract information from the candidate set and construct a knowledge graph with a four-tuple structure. , R represents an entity, R represents a relation, T represents a time attribute, and F represents a fuzzy rule; S6. Using the quadruple structure G as input, generate temporal fuzzy rules through the recursive reconstruction contribution method; S7. Incrementally update the knowledge graph based on temporal fuzzy rules.

[0004] Furthermore, the electrical performance parameters of the wafer, derived from impedance spectroscopy analysis, are expressed as follows:

[0005] in, and Let represent the amplitude and phase of the k-th frequency domain component of the impedance spectrum, respectively. K represents the total number of components, j represents the imaginary number, Z represents the complex impedance, and Z represents the electrical impedance characteristics of the wafer at a specific frequency.

[0006] Furthermore, the alignment of features X, Y, and Z is achieved through temporal contrastive loss, as shown in the formula:

[0007] in, Indicates time-series comparison loss. Indicates sample pairs and P represents a pair of positive feature samples, and N represents a pair of negative feature samples. Indicates sample pairs and , for Embedded representation, for Embedded representation, for Embedded representation, This represents the i-th feature in the feature set X. This represents the j-th feature in the feature set Y. This represents the k-th feature in the feature set Z.

[0008] Furthermore, the features are enhanced using the following formula:

[0009]

[0010]

[0011] in, Indicates enhanced , Let 'a' represent element-wise multiplication, 'a' represent attention weights, and 'W' represent the fuzzy feature correlation matrix. This represents the activation degree of the j-th rule. Representation rules The set of conditions, This represents the j-th rule. Represents the i-th variable Value The membership function.

[0012] Furthermore, the formula for the weighted scoring model is as follows:

[0013] in, This represents the weighted score of the i-th paragraph. This represents the i-th paragraph. Let F represent the i-th fuzzy feature, and let F represent the set of fuzzy features. Representing fuzzy features The weight, Representing fuzzy features In paragraph TF-IDF value in Representing time series features The weight, Represents the k-th time series feature. Representing time series features With paragraph The semantic matching degree, where T represents the temporal feature set.

[0014] Furthermore, time attributes are extracted from the text, and the degree of violation of time attributes is quantified using a dynamic membership function, which is expressed as:

[0015] in, Indicates that t represents time. This represents the tolerance time window, where δ is the tolerance coefficient.

[0016] Furthermore, S6 specifically refers to: The time-series contribution of variables to the fault index is calculated using the following formula:

[0017] in, Representing variables The time-series contribution of the failure index Y Indicates from variable The set of paths to the failure index Y considering timing constraints, where p represents... In the context of a path, let 'e' represent an edge on path 'p'. and These represent the edge weights and membership degrees considering time constraints, respectively. Define the membership function:

[0018] in, Representing variables The membership function, where x represents the variable. The value of k and This represents the parameter that controls the shape of the membership function. Representing time variables The membership function, where t represents time. and Representing time variables respectively The start and end times, τ and τ are the tolerance parameters for the time window; Select key variables whose time-series contribution is greater than or equal to a set threshold, and generate time-series fuzzy rules; The activation degree of temporal fuzzy rules is calculated jointly by parameters and temporal membership:

[0019] in, This represents the activation degree of the j-th temporal fuzzy rule. Representation rules The set of variables included in the conditions. Representation rules The condition includes a set of time variables, where xi represents the variable. The value, Representing time variables The value of .

[0020] The present invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for constructing a fuzzy knowledge graph regarding wafer manufacturing processes.

[0021] The present invention also proposes an electronic device including a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program including computer-readable instructions, and the processor is configured to invoke the computer-readable instructions to execute the above-described method for constructing a fuzzy knowledge graph regarding wafer manufacturing processes.

[0022] The present invention also proposes a computer program product, including a computer program / instruction that, when executed by a processor, implements the steps of the above-described method for constructing a fuzzy knowledge graph regarding wafer manufacturing processes.

[0023] The beneficial effects of the technical solution provided by this invention are: This invention extracts multimodal data features from the wafer manufacturing process, guides feature weighting based on fuzzy knowledge, constructs a candidate set of temporal fuzzy segments by weighting and scoring the fuzzy feature set and the temporal-series feature set, and extracts spatiotemporal constraint knowledge from the candidate set through a dynamic membership function to quantify the degree of constraint violation, thus enabling the coordinated expression of the time dimension and fuzzy logic. Based on the quadruplet structure, a recursive reconstruction contribution method is used to generate temporal fuzzy rules, and the knowledge graph is incrementally updated according to the temporal fuzzy rules. This invention breaks through the bottleneck of traditional methods in representing the spatiotemporal coupling and uncertainty correlation between wafer manufacturing defects and process parameters. Attached Figure Description

[0024] Figure 1 This is a flowchart of a method for constructing a fuzzy knowledge graph related to wafer manufacturing processes, according to an embodiment of the present invention. Figure 2 This is a block diagram of an electronic device according to an exemplary embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0026] The flowchart of the fuzzy knowledge graph construction method for wafer manufacturing processes in this embodiment of the invention is as follows: Figure 1 Specifically, it includes the following steps: S1. Acquire scanning electron microscope images, sensor timing data, parameter settings, and text of the rule base during the wafer manufacturing process.

[0027] Scanning electron microscopy (SEM) uses a line-by-line, point-by-point imaging method to convert different information features of the sample surface into video signals in a sequential and proportional manner. Finally, a magnified microscopic image of the sample surface corresponding to the image is observed on a fluorescent screen. This allows for in-depth observation of tiny areas and the acquisition of high-resolution surface morphology images.

[0028] Sensor timing data, such as power sensors and impedance analyzers, can be used to obtain information on plasma etching uniformity; pressure sensors can be used to obtain thin film deposition rate drift; thermocouples / infrared temperature measurement can be used to obtain photoresist baking stress deformation; and displacement encoders can be used to obtain photolithography alignment accuracy deviation.

[0029] The rule base includes: (1) Process parameter control rules, such as beam current density exceeding the limit by ±2% during ion implantation → automatic calibration of the magnetic lens. (2) Defect generation prevention and control rules, such as etching rate decreasing by 10%, sidewall angle < 88°, short circuit probability increasing by 40% → activation of AOI (Automated Optical Inspection) full inspection. (3) Equipment interaction rules, such as etching machine RF power supply reflected power > incident power × 15% → power reduced to a safe value. (4) Material interaction rules, such as thin film combination Cu / low k dielectric is prohibited from use. Plasma ashing, changed to Mixed gas. (5) Quality control rules. (6) Expert experience rule base, such as: when the following occurs: OES spectral fluctuation index > 0.4; impedance matching times > 20 times / min; wafer edge temperature difference > 3℃, execute: 1. Confidence level = 70% → reduce power by 10% and observe; 2. Confidence level > 90% → inject cleaning gas.

[0030] S2. A three-dimensional convolutional network is used to extract spatial defect features from scanning electron microscope images to obtain feature set X. The sensor time series data is encoded by a bidirectional LSTM to obtain feature set Y. The electrical performance parameters of the wafer based on impedance spectrum analysis are obtained from the text to obtain feature set Z.

[0031] Based on impedance spectroscopy analysis of the electrical performance parameters of the wafer, a quantitative correlation between dielectric properties and defects is established, expressed as:

[0032] in, and Let represent the amplitude and phase of the k-th frequency domain component of the impedance spectrum, respectively. K represents the total number of components, j represents the imaginary number, Z represents the complex impedance, and Z represents the electrical impedance characteristics of the wafer at a specific frequency.

[0033] Construct a unified embedding space for features X, Y, and Z Where d is the embedding dimension, The embedded representation of X, This represents the embedded representation of Y. This represents the embedding representation of Z.

[0034] Alignment of features X, Y, and Z is achieved through temporal contrastive loss, as shown in the formula:

[0035] in, Indicates time-series comparison loss. Indicates sample pairs and P represents a positive feature pair (multimodal data from the same wafer), and N represents a negative feature pair (multimodal data from different wafers). Indicates sample pairs and , for Embedded representation, for Embedded representation, for Embedded representation, This represents the i-th feature in the feature set X. This represents the j-th feature in the feature set Y. This represents the k-th feature in the feature set Z.

[0036] S3. Strengthen feature sets X, Y, and Z by constructing fuzzy knowledge based on text. This involves pre-constructing a defect rule base based on text. (Contains m fuzzy rules), define the fuzzy feature correlation matrix. The following formula is used to enhance the features, focusing on amplifying process parameter features that are highly correlated with the current defect mode.

[0037]

[0038]

[0039]

[0040] in, Indicates enhanced , Let 'a' represent element-wise multiplication, 'a' represent attention weights, and 'W' represent the fuzzy feature correlation matrix. This represents the activation degree of the j-th rule. Representation rules The set of conditions, This represents the j-th rule. Represents the i-th variable Value In this embodiment of the invention, the variable refers to the process parameter variable, such as chemical process parameters: gas flow rate ratio, concentration parameter, etching rate, etc. This represents the "etching rate" variable. =150Å / min is a specific value. This indicates the membership degree of the etching rate value "150 Å / min" in the fuzzy set of the "normal range".

[0041] S4. Construct a fuzzy feature set based on the enhanced feature sets X, Y and Z, extract a temporal feature set from the enhanced feature set Y, construct a weighted scoring model based on the fuzzy feature set and the temporal feature set, and select a candidate set from the text according to the weighted scoring model.

[0042] The fuzzy feature set F is obtained by applying the enhanced feature φ e Semantic mapping is constructed. The specific process is as follows: Important features are selected based on the attention weight 'a', and then numerical features are converted into fuzzy semantic descriptions. The construction formula is: , Where F is the fuzzy feature set. For the j-th fuzzy feature, such as "etching rate is significantly low", [i] represents the value of the i-th enhanced feature, and a[i] represents the attention weight corresponding to the i-th feature. The threshold for fuzzy features is used to filter out low-importance features, and the semantic_map function converts numerical features into fuzzy descriptions.

[0043] The temporal feature set T is specifically derived from enhanced temporal data. Extracting time patterns. The construction process includes time window identification and temporal constraint extraction:

[0044]

[0045] Where T is the set of time-series features. For the k-th temporal feature, such as "2-4 hours after deposition", For the enhanced feature portion from sensor time-series data Y, The time constraints are extracted from the rule base. Temporal characteristics The weights are defined as follows: Y_indices is the set of feature indices belonging to the time series data Y, |Y_indices| is the size of the time series feature index set, extract_temporal is the time series pattern extraction function, time_window is the time window constraint extraction function, and rules are the time constraint rules in the rule base.

[0046] Existing knowledge extraction methods struggle to simultaneously address two key characteristics of wafer manufacturing: the strict temporal constraints of process steps (e.g., photolithography must be completed within a specific time window after cleaning) and the fuzziness of parameter relationships (e.g., the uncertainty quantification in "high temperature may cause defects"). Traditional methods typically neglect the synergistic expression of the time dimension and fuzzy logic, limiting the applicability of knowledge graphs in dynamic reasoning scenarios.

[0047] Establish a dual-feature collaborative screening mechanism and define a fuzzy feature set. (e.g., "significantly correlated" or "may induce") and time series feature sets (e.g., "after" or "lasting [a,b] hours").

[0048] The formula for the weighted scoring model is as follows:

[0049] in, This represents the weighted score of the i-th paragraph. This represents the i-th paragraph. Let F represent the i-th fuzzy feature, and let F represent the set of fuzzy features. Representing fuzzy features The weight, Representing fuzzy features In paragraph TF-IDF value in Representing time series features The weight, Represents the k-th time series feature. Representing time series features With paragraph The semantic matching degree, where T represents the temporal feature set.

[0050] Segments with scores above the threshold τ (such as "abnormal etching rate within 2 hours after deposition will increase the risk of wafer warping") are selected to construct a candidate set.

[0051] S5. Extract information from the candidate set, including process entities and their spatiotemporal attributes, and construct a knowledge graph with a four-tuple structure. , Let ε represent the entity, R represent the relation, T represent the time attribute (recording the start / end time of the process step), and F represent the fuzzy rule. A fuzzy rule is defined as "If the etching rate is less than 150 Å / min and the time interval between deposition and etching is greater than 4 hours, then the film defect risk ≥ 0.7". For example, a quadruple G = (etching process, influence, [0-4 hours after deposition], IF etching rate < 150 Å / min AND time interval > 4 hours THEN film defect risk ≥ 0.7). Where: ε = "etching process" (entity); R = "influence" (relationship); T = "0-4 hours after deposition" (time attribute, recording the time constraint of the process step); F = "IF etching rate < 150 Å / min AND time interval > 4 hours THEN film defect risk ≥ 0.7" (fuzzy rule).

[0052] The method for identifying the spatiotemporal attributes of entities is as follows: Extract the time attribute. For example, for statements like "Etching needs to be completed ∆t∈[0,4] hours after deposition", quantify the degree of violation of the time attribute using a dynamic membership function. The dynamic membership function is expressed as:

[0053] in, Indicates that t represents time. This represents the tolerance time window, where δ is the tolerance coefficient.

[0054] S6. Using the quadruple structure G as input, generate temporal fuzzy rules through a recursive reconstruction contribution method.

[0055] First, calculate the time-series contribution of variables to the failure index (such as "defect density"), using the following formula:

[0056] in, Representing variables The time-series contribution of the failure index Y Indicates from variable The set of paths to the failure index Y considering timing constraints, where p represents... In the context of a path, let 'e' represent an edge on path 'p'. and These represent the edge weights and membership degrees considering time constraints (e.g., "the weight decay factor of the etching rate within t∈[0,4] hours after deposition is 0.8").

[0057] Secondly, timing conditions are explicitly introduced when generating rules. For variables... (e.g., etching rate) and time variables (e.g., "distance from deposition time"), define the membership function:

[0058] in, Representing variables The membership function, where x represents the variable. The value of k and This represents the parameter that controls the shape of the membership function. Representing time variables The membership function, where t represents time. and Representing time variables respectively The start and end times, τ and τ are the tolerance parameters for the time window; Key variables with a temporal contribution greater than or equal to a set threshold are selected to generate temporal fuzzy rules. For example, a rule that states "if the etching rate is between [100, 150] Å / min and the time interval between deposition and etching is between [0, 4] h, then the fault type is thin film defect risk".

[0059] The activation degree of temporal fuzzy rules is calculated jointly by parameters and temporal membership:

[0060] in, This represents the activation degree of the j-th temporal fuzzy rule. Representation rules The set of variables included in the conditions. Representation rules The set of time variables included in the conditions. Representing variables The value, Representing time variables The value of .

[0061] S7. Incrementally update the knowledge graph G based on temporal fuzzy rules.

[0062] The temporal fuzzy rules generated by S6 are integrated with the quadruple structure G to construct a complete fuzzy knowledge graph. The specific process includes: first, using the quadruple structure... The basic nodes and edges are defined as follows: entity ε is the graph node, relation R is the directed edge between nodes, time attribute T is the time weight of the edge, and fuzzy rule F is the confidence weight of the edge. Then, the temporal fuzzy rules are sorted and filtered according to their activation degree. The activation degree calculation formula is:

[0063] in, Let represent the activation degree of the j-th temporal fuzzy rule. Representation rules The set of variables included in the conditions. Representation rules The set of time variables included in the conditions. Representing variables Value Membership function at time, Representing time variables Value The membership function at that time.

[0064] When constructing the topology of the graph, connections are established through the relationships R between entities, and the condition-conclusion pairs in the temporal fuzzy rules are mapped to paths in the graph. For each rule, the variables and time constraints in its conditions are converted into paths in the graph from the premise entity to the conclusion entity. Finally, the weights of each node and edge in the graph are calculated. Node weights are determined based on the frequency of the entity's occurrence in all rules, and edge weights combine the relationship strength and the activation degree of the corresponding fuzzy rule. Computation. The completed knowledge graph can support defect prediction and process optimization decisions based on fuzzy reasoning, and realize the effective characterization of spatiotemporal coupling relationships and the quantitative analysis of uncertainty correlations in the wafer manufacturing process.

[0065] In one exemplary embodiment, a computer-readable storage medium is included, which stores a computer program that, when executed by a processor, implements the above-described method for constructing a fuzzy knowledge graph regarding wafer manufacturing processes.

[0066] Please see Figure 2 In one exemplary embodiment, the device further includes an electronic device including at least one processor, at least one memory, and at least one communication bus.

[0067] The memory stores a computer program, which includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through a communication bus to execute the aforementioned fuzzy knowledge graph construction method for wafer manufacturing processes.

[0068] In one exemplary embodiment, a computer program product is proposed, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method for constructing a fuzzy knowledge graph regarding wafer manufacturing processes.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a fuzzy knowledge graph about wafer manufacturing processes, characterized in that, The following steps are involved: S1. Acquire scanning electron microscope images, sensor timing data, parameter settings, and text of the rule base during the wafer manufacturing process; S2. A three-dimensional convolutional network is used to extract spatial defect features from scanning electron microscope images to obtain feature set X. The sensor time series data is encoded by a bidirectional LSTM to obtain feature set Y. The electrical performance parameters of the wafer based on impedance spectrum analysis are obtained from the text to obtain feature set Z. S3. Strengthen feature sets X, Y, and Z by constructing fuzzy knowledge based on text; S4. Construct a fuzzy feature set based on the enhanced feature sets X, Y and Z, extract a temporal feature set from the enhanced feature set Y, construct a weighted scoring model based on the fuzzy feature set and the temporal feature set, and filter candidate sets from the text according to the weighted scoring model; S5. Extract information from the candidate set and construct a knowledge graph with a four-tuple structure. , R represents an entity, R represents a relation, T represents a time attribute, and F represents a fuzzy rule; S6. Using the quadruple structure G as input, generate temporal fuzzy rules through the recursive reconstruction contribution method; S7. Incrementally update the knowledge graph based on temporal fuzzy rules.

2. The method for constructing a fuzzy knowledge graph about wafer manufacturing processes according to claim 1, characterized in that, The electrical performance parameters of the wafer, derived from impedance spectroscopy analysis, are expressed as follows: in, and Let represent the amplitude and phase of the k-th frequency domain component of the impedance spectrum, respectively. K represents the total number of components, j represents the imaginary number, Z represents the complex impedance, and Z represents the electrical impedance characteristics of the wafer at a specific frequency.

3. The method for constructing a fuzzy knowledge graph about wafer manufacturing processes according to claim 1, characterized in that, Alignment of features X, Y, and Z is achieved through temporal contrastive loss, as shown in the formula: in, Indicates time-series comparison loss. Indicates sample pairs and P represents a pair of positive feature samples, and N represents a pair of negative feature samples. Indicates sample pairs and , for Embedded representation, for Embedded representation, for Embedded representation, This represents the i-th feature in the feature set X. This represents the j-th feature in the feature set Y. This represents the k-th feature in the feature set Z.

4. The method for constructing a fuzzy knowledge graph about wafer manufacturing processes according to claim 1, characterized in that, The feature is enhanced using the following formula: in, Indicates enhanced , Let 'a' represent element-wise multiplication, 'a' represent attention weights, and 'W' represent the fuzzy feature correlation matrix. This represents the activation degree of the j-th rule. Representation rules The set of conditions, This represents the j-th rule. Represents the i-th variable Value The membership function.

5. The method for constructing a fuzzy knowledge graph about wafer manufacturing processes according to claim 1, characterized in that, The formula for the weighted scoring model is as follows: in, This represents the weighted score of the i-th paragraph. This represents the i-th paragraph. Let F represent the i-th fuzzy feature, and let F represent the set of fuzzy features. Representing fuzzy features The weight, Representing fuzzy features In paragraph TF-IDF value in Representing time series features The weight, Represents the k-th time series feature. Representing time series features With paragraph The semantic matching degree, where T represents the temporal feature set.

6. The method for constructing a fuzzy knowledge graph about wafer manufacturing processes according to claim 1, characterized in that, The time attribute is extracted from the text, and the degree of violation of the time attribute is quantified by a dynamic membership function, which is expressed as: in, Indicates that t represents time. This represents the tolerance time window, where δ is the tolerance coefficient.

7. The method for constructing a fuzzy knowledge graph about wafer manufacturing processes according to claim 1, characterized in that, S6 specifically refers to: The time-series contribution of variables to the fault index is calculated using the following formula: in, Representing variables The time-series contribution of the failure index Y Indicates from variable The set of paths to the failure index Y considering timing constraints, where p represents... In the context of a path, let 'e' represent an edge on path 'p'. and These represent the edge weights and membership degrees considering time constraints, respectively. Define the membership function: in, Representing variables The membership function, where x represents the variable. The value of k and This represents the parameter that controls the shape of the membership function. Representing time variables The membership function, where t represents time. and Representing time variables respectively The start and end times, τ and τ are the tolerance parameters for the time window; Select key variables whose time-series contribution is greater than or equal to a set threshold, and generate time-series fuzzy rules; The activation degree of temporal fuzzy rules is calculated jointly by parameters and temporal membership: in, This represents the activation degree of the j-th temporal fuzzy rule. Representation rules The set of variables included in the conditions. Representation rules The condition includes a set of time variables, where xi represents the variable. The value, Representing time variables The value of .

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the method as described in any one of claims 1-7.

9. An electronic device, characterized in that, The device includes a processor and a memory interconnected thereto, wherein the memory is used to store a computer program, the computer program including computer-readable instructions, and the processor is configured to invoke the computer-readable instructions to perform the method as described in any one of claims 1-7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-7.

Images