Parasitic parameter result automatic analysis system based on FUZZ test

Abstract

The present application relates to the technical field of integrated circuit design automation data processing, and especially relates to a parasitic parameter result automatic analysis system based on FUZZ testing, which analyzes original files and generates standardized data through a data preprocessing module, a fuzzy testing engine module generates a test case library including illegal characters, extreme values and logic loops by adopting a syntax tree variation, an abnormal response module captures illegal operations in the analysis and simulation stages through a probe detection unit, a correction decision unit executes data repair or generates an abnormal report, and an incremental update unit synchronizes abnormal features to a model self-optimization module rule library. A dynamic adaptation module converts multi-process parasitic parameters into a standardized simulation model through a process parameter abstraction layer. The modules form a closed-loop system of test case generation, abnormality detection and repair, and rule dynamic optimization, realize automatic processing of unexpected input scenarios, and improve the robustness and multi-process compatibility of parasitic parameter analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of integrated circuit design automation data processing, and particularly relates to a parasitic parameter result automatic analysis system based on FUZZ testing. BACKGROUND

[0002] In the field of integrated circuit design, parasitic parameter extraction is a key link of circuit performance verification. After the existing RC tool extracts parasitic parameters, it usually relies on manual operation for data analysis and model optimization, which is difficult to cope with complex and variable abnormal data scenarios. In the prior art, the parasitic parameter extraction system generally lacks a robust testing mechanism for unexpected input data, especially when dealing with format error, boundary value abnormality or logic conflict of parasitic parameter data, the system is prone to analysis interruption or error output, resulting in prolonged design iteration cycle. For example, when the parasitic parameter file generated by the RC tool contains illegal characters or out-of-range values, the existing analysis system cannot automatically identify the abnormality and trigger the correction process, and manual intervention is required to troubleshoot the data source problem, which not only reduces the verification efficiency, but also causes model parameter mismatch due to delayed response. In addition, the conventional parameterized interface cannot dynamically adapt to the data characteristics of different process nodes, and it is difficult to maintain stable processing capacity when encountering non-standardized input, which directly affects the subsequent simulation accuracy. SUMMARY

[0003] In view of the deficiencies in the prior art, the present application provides a parasitic parameter result automatic analysis system based on FUZZ testing, which solves the problem that the existing parasitic parameter analysis system lacks automatic abnormal data processing mechanism and dynamic adaptation capability, resulting in insufficient analysis robustness in unexpected input scenarios, causing simulation accuracy to decline and design iteration efficiency to decrease.

[0004] To solve the above technical problems, the specific technical solutions of the present application are as follows:

[0005] The parasitic parameter result automatic analysis system based on FUZZ testing provided by the present application comprises: a data preprocessing module, a fuzzy testing engine module, a parameterized interface module, an abnormal response module, a visual analysis module, a model self-optimization module and a dynamic adaptation module.

[0006] The data preprocessing module is used to receive the original file output by the parasitic parameter extraction tool, and extracts the capacitance, resistance network topology and process node attribute field through line-by-line scanning by a syntax analysis unit. After verifying the node connection relationship and value range by a structure verification unit, the data cleaning unit performs unit conversion and interpolation completion, and outputs standardized data.

[0007] The fuzzy test engine module receives standardized data, generates a test case library including illegal characters, extreme values and logical loops by syntax tree mutation of numerical fields and topological connectors through a mutation case generation unit, a stress test unit injects the mutated cases into an analysis process and monitors system crash points, a vulnerability mapping unit associates error types with trigger conditions to generate a vulnerability mapping table, and pushes it to the abnormal response module;

[0008] The parameterized interface module dynamically loads YAML format analysis rule sets based on process node keywords, converts mixed delimiters through intelligent identification and matching of multi-path parsers, generates intermediate representation format data and outputs it to the abnormal response module and the visual analysis module;

[0009] The abnormal response module embeds a probe detection unit in the analysis and simulation stages according to the vulnerability mapping table to capture illegal operations, calls an interpolation algorithm to complete repairable abnormal data through a correction decision unit, generates an error code containing abnormal report and pushes it to the parasitic parameter extraction tool to terminate the task, and an incremental update unit synchronizes the abnormal scene features to the analysis rule library of the parameterized interface module;

[0010] The visual analysis module receives intermediate representation format data and abnormal reports, maps the parasitic parameter network through a topological graph rendering engine, superimposes abnormal area heat maps and scatter plots, and generates layout optimization suggestions to feed back to the model self-optimization module;

[0011] The model self-optimization module uses unsupervised learning to cluster abnormal scenes, extracts high-frequency abnormal patterns to generate new analysis rules, distributes them to the parameterized interface module through a message queue to adjust the fault tolerance threshold, and synchronizes them to the dynamic adaptation module;

[0012] The dynamic adaptation module matches process feature words in the process keyword library, loads corresponding analysis algorithms through a strategy switching unit, and converts different process parasitic parameters into standardized simulation models and adapts to SPICE simulation.

[0013] Further, the FUZZ test-based parasitic parameter result automatic analysis system of the present application comprises:

[0014] The syntax analysis unit receives the original file output by the parasitic parameter extraction tool, extracts the capacitance, resistance network topology and process node attribute field through the syntax analysis function of the data preprocessing module by using regular expression and finite state machine line-by-line scanning, and outputs to the structure verification unit; the structure verification unit verifies whether the node connection relationship and numerical range meet the process node definition based on the YAML format analysis rule set dynamically loaded by the parameterized interface module, generates a verification log to mark the abnormal field and passes it to the data cleaning unit; the data cleaning unit performs unit uniform conversion and interpolation completion operation on the data output by the structure verification unit to generate a standardized data table and output to the fuzzy test engine module.

[0015] Further, the FUZZ test-based parasitic parameter result automatic analysis system of the application comprises:

[0016] The mutation case generation unit receives the standardized data output by the data preprocessing module, disturbs the numerical field and topology connection symbol based on the syntax tree mutation function of the fuzzy test engine module, generates a test case library including illegal characters, extreme values and logical loops, and outputs to the stress test unit; the stress test unit injects the mutation case into the analysis process of the parameterized interface module through the data pipeline defined by the fuzzy test engine module, monitors the system crash point and timeout state, records the error type and trigger condition and passes it to the vulnerability mapping unit; the vulnerability mapping unit associates the test results recorded by the stress test unit with the mutation case characteristics generated by the fuzzy test engine module, constructs a mapping relationship table of vulnerability types and trigger conditions, and pushes it to the abnormal response module to update the detection rules.

[0017] Further, the FUZZ test-based parasitic parameter result automatic analysis system of the application comprises:

[0018] The rule loading unit dynamically loads the field separator, numerical precision and topology connection symbol definition in YAML format based on the analysis rule library distributed by the model self-optimization module through the message queue, and outputs to the intelligent identification unit;

[0019] The intelligent identification unit receives the test case library generated by the fuzzy test engine module, converts the mixed separator to intermediate representation format data according to the multi-way parser defined by the parameterized interface module according to the process keywords, and outputs to the format compatible unit;

[0020] The format compatible unit processes the SPICE netlist, CSV matrix and custom binary format data in parallel, generates uniform intermediate representation format data and outputs to the abnormal response module and visual analysis module.

[0021] Further, the FUZZ test-based parasitic parameter result automatic analysis system of the present application, the abnormal response module comprises: a probe detection unit, a probe is embedded in the analysis process and the simulation stage of the parameterized interface module, based on the vulnerability mapping table generated by the fuzzy test engine module, real-time capture of zero error and memory overflow operation, and the detection result is transmitted to the correction decision unit; the correction decision unit, the interpolating algorithm of the data preprocessing module is called to complete the data for the repairable abnormality, and the analysis process of the parameterized interface module is rewritten to reanalyze, and the report with the error code is generated for the unrepairable abnormality and pushed to the parasitic parameter extraction tool to terminate the task; the incremental update unit adds the abnormal scene characteristics captured by the probe detection unit to the rule library of the model self-optimization module, adjusts the mutation case generation weight of the fuzzy test engine module based on the clustering analysis result of the model self-optimization module, and synchronizes to the parameterized interface module through the message queue.

[0022] Further, the FUZZ test-based parasitic parameter result automatic analysis system of the present application, the model self-optimization module comprises: a clustering analysis unit, receiving abnormal scene data detected by the abnormal response module, based on the layout optimization suggestion generated by the visual analysis module, using DBSCAN algorithm in unsupervised learning to cluster metal layer capacitance coupling characteristics and high frequency abnormal mode, outputting the clustering result to the rule update unit;

[0023] The rule update unit generates new analysis rules according to the clustering result of the clustering analysis unit, and adjusts the fault tolerance threshold based on the vulnerability mapping table of the fuzzy test engine module, and distributes to the parameterized interface module and the fuzzy test engine module through the message queue;

[0024] The feedback loop unit synchronizes the updated analysis rules and fault tolerance threshold to the dynamic adaptation module through the message queue of the model self-optimization module, and triggers the visual analysis module to re-render the abnormal area heat map.

[0025] Further, the FUZZ test-based parasitic parameter result automatic analysis system of the present application, the dynamic adaptation module comprises:

[0026] The keyword matching unit receives the uniform intermediate representation format data generated by the parameterized interface module, extracts the process feature words and matches the mapping library of the process node and the analysis algorithm based on the process keyword library synchronized by the model self-optimization module, and outputs to the strategy switching unit;

[0027] The strategy switching unit deploys the lightweight semantic analyzer defined by the dynamic adaptation module, dynamically loads the updated analysis algorithm and verification rule of the parameterized interface module, and outputs to the simulation conversion unit;

[0028] The simulation conversion unit converts parasitic parameters of different processes into a standardized simulation model through a process parameter abstraction layer of the dynamic adaptation module, outputs to a visual analysis module to render an abnormal area thermodynamic diagram, and adapts to SPICE simulation through a parameterized interface module.

[0029] The present application has the following advantages.

[0030] The present application significantly improves the robustness and design iteration efficiency of parasitic parameter analysis by constructing an automatic abnormality processing and dynamic adaptation closed-loop system. The data preprocessing module uses regular expressions and finite state machines to analyze multi-source heterogeneous data, generates standardized input by combining structure checking and cleaning mechanism, and eliminates original data noise and format differences; the fuzzy test engine generates a test case library covering unexpected input based on syntax tree variation, establishes an abnormal trigger rule library through stress testing and vulnerability mapping, and enhances the detection capability of the system for illegal characters, extreme values and logical conflicts. The abnormal response module captures and analyzes abnormal operations in the simulation stage in real time through the probe, and the decision unit realizes data repair and task control based on interpolation algorithm, and the incremental update mechanism synchronizes the abnormal characteristics to the model self-optimization module to drive rule iteration; the parameterized interface module dynamically loads and analyzes the rule set, and realizes multi-process simulation model conversion combined with the process parameter abstraction layer, and adapts to SPICE simulation verification. The visual analysis module feeds back the abnormal distribution through the thermodynamic diagram and topological diagram, the model self-optimization module optimizes the analysis rules and fault tolerance threshold by using unsupervised learning, the dynamic adaptation module switches strategies according to process characteristics, forms a closed-loop link of testing, optimization and feedback, realizes automatic processing of abnormal scenarios, multi-process compatible analysis and system self-healing capability, reduces the cost of manual intervention, and improves the precision and efficiency of integrated circuit design verification. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, for those skilled in the art, other drawings can also be obtained from the drawings without creative labor.

[0032] Figure 1 The system architecture diagram of the parasitic parameter result automatic analysis system based on FUZZ test provided by the embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of the present application. The technical solutions provided by each embodiment of the present application will be described in detail below in combination with the drawings. In order to better understand the objects of the present application, the present application will be further described in detail below.

[0034] Please refer to Figure 1 The present application provides a parasitic parameter result automatic analysis system based on FUZZ test, comprising: a data preprocessing module, a fuzzy test engine module, a parameterization interface module, an abnormal response module, a visual analysis module, a model self-optimization module and a dynamic adaptation module.

[0035] The data preprocessing module is used for receiving original files output by a parasitic parameter extraction tool, extracting capacitance and resistance network topology and process node attribute fields through line-by-line scanning by a syntax analysis unit, verifying node connection relationship and value range by a structure verification unit, and then performing unit conversion and interpolation completion by a data cleaning unit, and outputting standardized data.

[0036] The fuzzy test engine module receives the standardized data, performs syntax tree mutation on value fields and topology connection symbols by a mutation case generation unit, generates a test case library including illegal characters, extreme values and logic loops, injects the mutated cases into an analysis process by a stress test unit and monitors system crash points, generates a vulnerability mapping table by associating error types and trigger conditions by a vulnerability mapping unit, and pushes to the abnormal response module.

[0037] The parameterization interface module dynamically loads YAML format analysis rule sets based on process node keywords, converts mixed delimiters by matching a multi-way parser through an intelligent recognition unit, generates intermediate representation format data and outputs to the abnormal response module and the visual analysis module.

[0038] The abnormal response module embeds a probe detection unit in the analysis and simulation stages according to the vulnerability mapping table to capture illegal operations, calls an interpolation algorithm to complete repairable abnormal data by a correction decision unit, generates an exception report with error codes, pushes to the parasitic parameter extraction tool to terminate the task, and synchronizes the abnormal scene features to the analysis rule library of the parameterization interface module by an incremental update unit.

[0039] The visual analysis module receives the intermediate representation format data and the exception report, maps the parasitic parameter network by a topology graph rendering engine, superimposes the abnormal area heat map and scatter plot, and generates layout optimization suggestions to feed back to the model self-optimization module.

[0040] The model self-optimization module adopts unsupervised learning to cluster abnormal scenarios, extracts high-frequency abnormal patterns to generate new analysis rules, distributes the new analysis rules to the parameterization interface module through a message queue to adjust the fault tolerance threshold, and synchronizes the fault tolerance threshold to the dynamic adaptation module;

[0041] The dynamic adaptation module matches process characteristic words in a process keyword library, loads corresponding analysis algorithms through a strategy switching unit, converts different process parasitic parameters into standardized simulation models, and adapts SPICE simulation.

[0042] The FUZZ test-based parasitic parameter result automatic analysis system provided by the application realizes automatic processing and optimization of parasitic parameter data through multi-module cooperation. The data preprocessing module receives the original file output by the parasitic parameter extraction tool, and extracts the capacitance, resistance network topology and process node attribute field by line scanning through the syntax analysis unit. The syntax analysis unit uses regular expressions and finite state machines to analyze the data format, identifies key fields and extracts numerical parameters. The structure verification unit verifies the logical consistency of the node connection relationship and whether the numerical range meets the process node definition based on the YAML format analysis rule set dynamically loaded by the parameterization interface module, generates a verification log to mark abnormal fields. The data cleaning unit performs unit uniform conversion and interpolation completion operation on the verified data, converts the capacitance value and resistance value into standard units, fills in the missing data points through interpolation algorithm, and outputs the standardized data to the fuzzy test engine module.

[0043] After receiving the standardized data, the fuzzy test engine module, the variation case generation unit performs syntax tree variation on the numerical field and topology connection symbol, and generates a test case library including illegal characters, extreme values and logical loops. The variation operation is based on the abstract syntax tree structure, and the numerical boundary and operator are randomly disturbed to generate variation cases covering unexpected input scenarios. The stress test unit injects the variation cases into the analysis process of the parameterization interface module through the data pipeline, monitors the crash points and timeout states of the system in the analysis and simulation stages, records the error types and triggering conditions. The vulnerability mapping unit associates the stress test results with the characteristics of the variation cases, constructs a mapping relationship table of vulnerability types and triggering conditions, and pushes the mapping relationship table to the abnormal response module to update the detection rules.

[0044] The parameterization interface module dynamically loads YAML format analysis rule sets based on process node keywords, and the rule loading unit extracts field separators, numerical precision and topology connection symbol definitions from the analysis rule library synchronized by the model self-optimization module. The intelligent identification unit receives the test case library generated by the fuzzy test engine module, matches multiple parsers according to process keywords, and converts mixed separators into intermediate representation format data. The format compatible unit processes SPICE netlist, CSV matrix and custom binary format data in parallel, generates uniform intermediate representation format data and outputs the uniform intermediate representation format data to the abnormal response module and the visual analysis module.

[0045] The exception response module embeds a probe detection unit in the analysis and simulation stage according to the vulnerability mapping table, and captures zero error and memory overflow operations in real time. The probe detection unit inserts a detection point in the analysis process of the parameterized interface module, identifies high-risk operations based on the mapping table, and outputs the detection results. The correction decision unit calls the interpolation algorithm of the data preprocessing module to complete the repairable abnormal data, and writes back to the analysis process of the parameterized interface module for reanalysis; generates an error code report for the unrepairable abnormality and pushes it to the parasitic parameter extraction tool to terminate the task. The incremental update unit adds the abnormal scene features to the rule library of the model self-optimization module, adjusts the mutation case generation weight of the fuzz test engine module, and synchronizes to the parameterized interface module through the message queue.

[0046] The visual analysis module receives intermediate representation format data and exception reports, and maps the parasitic parameter network into an interactive topology graph through a topology graph rendering engine. The topology graph rendering engine layouts nodes and edges based on a force-directed algorithm, the node size represents the capacitance or resistance value, and the edge color depth reflects the abnormal level. The abnormal area heat map and scatter plot are superimposed on the layout coordinate system, and the specific abnormal point coordinates and process layer distribution are labeled. The layout optimization suggestions are generated based on historical repair records and fed back to the model self-optimization module to guide the rule optimization direction.

[0047] The model self-optimization module uses unsupervised learning to cluster abnormal scenes, the clustering analysis unit receives abnormal scene data detected by the exception response module, identifies metal layer capacitance coupling features and high-frequency abnormal patterns based on the DBSCAN algorithm. The rule update unit generates new analysis rules according to the clustering results, adjusts the fault tolerance threshold based on the vulnerability mapping table of the fuzz test engine module, and distributes them to the parameterized interface module and the fuzz test engine module through the message queue. The feedback closed loop unit synchronizes the updated analysis rules and fault tolerance threshold to the dynamic adaptation module, triggers the visual analysis module to re-render the abnormal area heat map, forming a closed loop link of rule optimization and visual feedback.

[0048] The dynamic adaptation module matches the process feature words in the process keyword library, the keyword matching unit receives the intermediate representation format data generated by the parameterized interface module, extracts the process feature words and associates the analysis algorithm mapping library. The strategy switching unit deploys a lightweight semantic analyzer to dynamically load corresponding analysis algorithms and verification rules, and switches the analysis strategy according to the process feature words. The simulation conversion unit converts parasitic parameters of different processes into standardized simulation models through the process parameter abstraction layer, outputs to the visual analysis module to render the heat map, and adapts the SPICE simulation through the parameterized interface module, realizing multi-process compatible automatic simulation verification. Each module cooperates through data flow and message queue to build a complete closed loop system from data preprocessing, abnormal testing to dynamic optimization.

[0049] Specifically, the FUZZ test-based parasitic parameter result automatic analysis system of the present application, the data preprocessing module comprises:

[0050] The syntax analysis unit receives the original file output by the parasitic parameter extraction tool, and extracts the capacitance, resistance network topology and process node attribute field through syntax analysis function of the data preprocessing module, line-by-line scanning by using regular expression and finite state machine, and outputs to the structure verification unit; the structure verification unit verifies whether the node connection relationship and numerical range conform to the process node definition based on the YAML format analysis rule set dynamically loaded by the parameterized interface module, generates a verification log to mark the abnormal field and passes it to the data cleaning unit; the data cleaning unit performs unit uniform conversion and interpolation completion operation on the data output by the structure verification unit after verification, and generates a standardized data table output to the fuzzy test engine module.

[0051] The data preprocessing module receives the original file output by the parasitic parameter extraction tool, which is initially processed by the syntax analysis unit. The syntax analysis unit uses regular expression to match the regular mode of capacitance value, resistance network topology connection relationship and process node attribute field, and combines finite state machine to scan the original file line by line, identifies the field boundary and extracts the key parameters. The regular expression mode is dynamically configured according to the process node type, and the finite state machine manages the analysis process through the state transition table to realize the compatibility of different data formats. The extracted capacitance, resistance network topology and process node attribute field are output to the structure verification unit to form the preliminary structured data.

[0052] The structure verification unit performs logical verification on the data output by the syntax analysis unit based on the YAML format analysis rule set dynamically loaded by the parameterized interface module. The YAML rule set defines the allowed capacitance value range of the process node, the connection constraint of the resistance network topology and the node attribute association relationship, and the structure verification unit traverses the node connection relationship to check whether there is an undefined node or illegal connection, and verifies whether the numerical parameter exceeds the threshold defined by the process node. The verification process generates a verification log, marks the abnormal field and adds an error type identifier, and passes the verified data to the data cleaning unit.

[0053] The data cleaning unit performs unit uniform conversion and interpolation completion operation on the data output by the structure verification unit after verification. Unit uniform conversion converts capacitance value and resistance value into preset standard units, eliminating unit differences of different data sources; interpolation completion fills in missing values based on the numerical trend of adjacent data points using linear interpolation algorithm to generate continuous and complete time series data sequence. The cleaned data is classified and aggregated according to capacitance type and process node to generate a standardized data table output to the fuzzy test engine module, providing consistent input for subsequent fuzzy test. The processing units are connected in series through data flow to form a complete preprocessing link from original data analysis, logical verification to standardized output.

[0054] Specifically, the FUZZ test-based parasitic parameter result automatic analysis system of the present application, the fuzz test engine module comprises:

[0055] The mutation case generation unit receives the standardized data output by the data preprocessing module, disturbs the numerical field and the topological connector based on the syntax tree mutation function of the fuzz test engine module, generates a test case library including illegal characters, extreme values and logical loops, and outputs to the stress test unit; the stress test unit injects the mutation case into the parsing process of the parameterized interface module through the data pipeline defined by the fuzz test engine module, monitors the system crash point and the timeout state, records the error type and the triggering condition and delivers to the vulnerability mapping unit; the vulnerability mapping unit associates the test results recorded by the stress test unit with the mutation case features generated by the fuzz test engine module, constructs a mapping relationship table of vulnerability types and triggering conditions, and pushes to the abnormal response module to update the detection rules.

[0056] After the fuzz test engine module receives the standardized data output by the data preprocessing module, the mutation case generation unit performs syntax tree mutation operation on the numerical field and the topological connector. Syntax tree mutation is based on abstract syntax tree structure, randomly disturbs the boundary value of the numerical field, the operator and the logical relationship of the topological connector, and generates a test case library including abnormal scenarios such as illegal characters, extreme values beyond the process limit and self-closing loops. During the mutation process, the disturbance range of the numerical field is dynamically adjusted according to the parameter threshold defined by the process node, and the mutation of the topological connector is realized by replacing or deleting the key node, covering the unexpected input combination. The generated test case library is output to the stress test unit through the data pipeline, forming the input source of the fuzz test.

[0057] The stress test unit injects the mutation case into the parsing process of the parameterized interface module through the predefined data pipeline, drives the system to execute abnormal data processing. The data pipeline adopts an asynchronous communication mechanism, schedules the injection order of the mutation case according to the priority queue, and monitors the memory occupancy rate and thread state of the parsing process. The system crash point is located by stack tracking technology, and the register state and memory snapshot at the time of process abnormal termination are recorded; the timeout state is determined based on the timer threshold, and the interrupt signal is triggered to forcibly terminate the unresponsive task. The error type is coded according to the preset classification rule, including format parsing error, numerical overflow and logical conflict, the triggering condition is associated with the input features of the mutation case, and the error log is formed and delivered to the vulnerability mapping unit.

[0058] The vulnerability mapping unit receives the test results recorded by the stress test unit and the characteristics of the variation case, and constructs a mapping relationship table of vulnerability types and triggering conditions. The mapping relationship table is stored in a key-value pair structure, the key is the error type code, and the value is the variation case characteristic set that triggers the error type, including the disturbance field position, the variation operation type and the input parameter range. When updating the mapping table, the repeated triggering scenarios are merged based on the feature similarity, and the high-frequency error mode is extracted. The updated mapping table is pushed to the abnormal response module to drive the incremental optimization of the detection rule library, forming a closed-loop feedback of the fuzz test results and the abnormal response strategy. Each processing unit is connected through a standardized interface and a data stream to realize the automatic test link from case generation, stress injection to vulnerability analysis.

[0059] Specifically, the FUZZ test-based parasitic parameter result automatic analysis system of the application comprises a parameterized interface module:

[0060] A rule loading unit loads dynamically the field separator, numerical precision and topology connector definition in YAML format based on the analysis rule library distributed by the model self-optimization module, and outputs to the intelligent recognition unit;

[0061] An intelligent recognition unit receives the test case library generated by the fuzz test engine module, matches the multi-way parser defined by the parameterized interface module according to the process keywords, converts the mixed separator into intermediate representation format data, and outputs to the format compatible unit;

[0062] A format compatible unit processes the SPICE netlist, CSV matrix and custom binary format data in parallel, generates unified intermediate representation format data and outputs to the abnormal response module and the visual analysis module.

[0063] The parameterized interface module receives the analysis rule library distributed by the model self-optimization module through the message queue, and the rule loading unit dynamically loads the field separator, numerical precision and topology connector definition in YAML format. The YAML rule set is stored according to the process node, and defines the data analysis parameters in different process scenarios, including the regular expression pattern of the field separator, the decimal place limit of the numerical precision and the legal combination rule of the topology connector. During the dynamic loading process, the rule loading unit listens to the update event of the message queue, loads the newly added or modified analysis rules in real time, generates an analysis configuration parameter table and outputs to the intelligent recognition unit, supporting dynamic adaptation of multiple process nodes.

[0064] The intelligent recognition unit receives the test case library generated by the fuzz test engine module, and matches a multi-path parser based on a process keyword. The process keyword is extracted from the metadata of the test case, including a process node identifier, a process type, and a design rule version, and is matched with a multi-path parser tag defined by the parameterized interface module. The multi-path parser identifies the context semantics of the mixed delimiter based on the configuration parameters provided by the rule loading unit, and uniformly converts delimiters such as commas, spaces, or tabs into intermediate representation format delimiters. The converted data is reorganized in a row-column structure to generate an intermediate representation format data stream, which is output to the format compatible unit to eliminate the influence of input data format differences on subsequent processing.

[0065] The format compatible unit processes the SPICE netlist, the CSV matrix, and the custom binary format data in parallel to generate unified intermediate representation format data. When the SPICE netlist is parsed, the connection relationship of the circuit elements and the parasitic parameter values are extracted and converted into an intermediate representation of node-edge structure; when the CSV matrix is parsed, the row-column index is mapped into the intermediate data in the form of key-value pair; and the custom binary format is restored to a structured data table through a deserialization algorithm. Parallel processing uses a multi-thread scheduling mechanism to allocate independent parsing threads for different data formats, and uses a mutex to ensure data consistency. The generated intermediate representation format data is sorted by timestamp and output to the exception response module for real-time detection, and is synchronized to the visualization analysis module to render a topology graph, realizing unified processing and collaborative analysis of multi-source heterogeneous data. The units are connected through configuration parameter tables, data streams, and message queues to build a complete data processing link from rule loading, intelligent parsing to format conversion.

[0066] Specifically, the parasitic parameter result automatic analysis system based on FUZZ test of the application comprises: a probe detection unit, which is embedded in the analysis process and simulation stage of the parameterized interface module, based on the vulnerability mapping table generated by the fuzz test engine module, real-time captures the division by zero error and memory overflow operation, and transmits the detection result to the correction decision unit; the correction decision unit, which calls the interpolation algorithm of the data preprocessing module to complete the data for the repairable exception, and writes back to the analysis process of the parameterized interface module to reanalyze, generates a report with an error code for the unrepairable exception, and pushes it to the parasitic parameter extraction tool to terminate the task; and the incremental update unit, which adds the exception scene features captured by the probe detection unit to the rule library of the model self-optimization module, adjusts the mutation case generation weight of the fuzz test engine module based on the clustering analysis result of the model self-optimization module, and synchronizes to the parameterized interface module through the message queue.

[0067] The abnormal response module deploys a probe detection unit in the analysis process and simulation stage of the parameterized interface module, and captures abnormal operations in real time based on a vulnerability mapping table generated by the fuzzy test engine module. The probe detection unit inserts a probe hook at the entrance and exit of a key function of an analysis thread to monitor illegal operations such as division-by-zero errors and memory overflow; the vulnerability mapping table stores the key-value mapping relationship between error types and triggering conditions, and the probe identifies abnormalities by matching the current operation characteristics with the mapping table entries, records the abnormal type, stack information and context data, and encapsulates the detection results as event messages to the correction decision unit.

[0068] After receiving the abnormal event message of the probe detection unit, the correction decision unit processes according to the abnormal type. For repairable abnormalities, the interpolation algorithm of the data preprocessing module is called to complete the missing or damaged fields based on the numerical trend of adjacent data points, and the repaired data packet is written back to the analysis process of the parameterized interface module for re-analysis; for unrepairable abnormalities, a structured report is generated by extracting error codes, abnormal positions and triggering conditions, and is pushed to the parasitic parameter extraction tool through an API interface to terminate the current task and prevent error propagation. After the repaired data is re-analyzed, the verification result is fed back to the probe detection unit to update the detection strategy of the vulnerability mapping table.

[0069] The incremental update unit receives the abnormal scene characteristics captured by the probe detection unit, adds the abnormal type, triggering condition and repair result to the rule library of the model self-optimization module. The rule library is versioned and records the association between abnormal scenes and analysis rules; based on the clustering analysis result of the model self-optimization module, the variation case features corresponding to the high-frequency abnormal mode are extracted, the variation case generation weight of the fuzzy test engine module is adjusted, and the disturbance case associated with the high-frequency abnormality is generated preferentially. The updated weight parameters are synchronized to the parameterized interface module through the message queue, dynamically optimizing the loading priority of the analysis rule set, forming a closed-loop feedback link of abnormal detection, repair and rule optimization. Each unit is connected through an event-driven mechanism and a data pipeline to realize real-time processing of abnormal response and system self-optimization.

[0070] Specifically, the FUZZ test-based parasitic parameter result automatic analysis system of the present application, the model self-optimization module comprises: a clustering analysis unit that receives abnormal scene data detected by the abnormal response module, generates layout optimization suggestions based on the visual analysis module, and clusters metal layer capacitance coupling characteristics and high-frequency abnormal modes using the DBSCAN algorithm in unsupervised learning, and outputs the clustering result to the rule update unit;

[0071] The rule update unit generates new analysis rules based on the clustering result of the clustering analysis unit, adjusts the fault tolerance threshold based on the vulnerability mapping table of the fuzzy test engine module, and distributes them to the parameterized interface module and the fuzzy test engine module through the message queue;

[0072] The feedback closed loop unit synchronizes the updated parsing rule and the fault tolerance threshold to the dynamic adaptation module through the message queue of the model self-optimization module, and triggers the visual analysis module to re-render the abnormal area heat map.

[0073] The model self-optimization module receives the abnormal scene data detected by the abnormal response module and the layout optimization suggestion generated by the visual analysis module, and the clustering analysis unit adopts the DBSCAN algorithm in unsupervised learning to perform clustering analysis on the metal layer capacitive coupling characteristics and high-frequency abnormal patterns. The abnormal scene data includes the abnormal type, trigger condition and repair state captured by the probe detection unit, and the layout optimization suggestion includes the abnormal area coordinates in the layout coordinate system and the process layer distribution characteristics. The DBSCAN algorithm identifies high-density abnormal clusters and outliers based on the density clustering principle, taking the metal layer capacitive coupling strength and abnormal frequency as feature vectors, and outputs the clustering label and cluster center coordinates to the rule updating unit to provide data support for parsing rule optimization.

[0074] The rule updating unit generates a new parsing rule according to the clustering result of the clustering analysis unit, and adjusts the fault tolerance threshold based on the vulnerability mapping table of the fuzzy test engine module. The new parsing rule is generated based on the trigger condition of the high-frequency abnormal cluster, including the fault tolerance range of the field separator, the numerical accuracy tolerance interval and the replacement rule of the topological connector; the fault tolerance threshold is dynamically adjusted according to the abnormal trigger frequency recorded in the vulnerability mapping table, and the detection sensitivity of the high-frequency abnormal type is reduced to improve the parsing efficiency. The generated rule and threshold are distributed to the parameterized interface module to update the parsing rule set through the message queue, and are synchronized to the fuzzy test engine module to adjust the mutation case generation strategy, forming a dynamic adaptation mechanism of the rule library.

[0075] The feedback closed loop unit synchronizes the updated parsing rule and the fault tolerance threshold to the dynamic adaptation module through the message queue of the model self-optimization module, and drives the strategy switching of the process parameter abstraction layer. During the synchronization process, the message queue pushes the rule update event in the publish-subscribe mode, and after the dynamic adaptation module loads the new rule, the visual analysis module is triggered to re-render the abnormal area heat map. The heat map is color-coded based on the updated parsing rule for the abnormal area, and the coverage range of the optimized fault tolerance threshold is marked, forming a closed loop link of rule optimization and visual feedback. Through the cooperative mechanism of data flow and event-driven, the adaptive optimization process from abnormal clustering analysis to system global strategy adjustment is realized.

[0076] Specifically, the FUZZ test-based parasitic parameter result automatic analysis system of the application comprises:

[0077] The keyword matching unit receives the uniform intermediate representation format data generated by the parameterized interface module, extracts the process feature words based on the process keyword library synchronized by the model self-optimization module, matches the process node and the mapping library of the analysis algorithm, and outputs to the strategy switching unit.

[0078] The strategy switching unit deploys the lightweight semantic analyzer defined by the dynamic adaptation module, dynamically loads the updated analysis algorithm and verification rule of the parameterized interface module, and outputs to the simulation conversion unit.

[0079] The simulation conversion unit converts the parasitic parameters of different processes into standardized simulation models through the process parameter abstraction layer of the dynamic adaptation module, outputs to the visualization analysis module to render the abnormal area heat map, and adapts the SPICE simulation through the parameterized interface module.

[0080] The dynamic adaptation module receives the uniform intermediate representation format data generated by the parameterized interface module, and the keyword matching unit extracts the process feature words based on the process keyword library synchronized by the model self-optimization module. The process keyword library stores the feature word set of different process nodes, including process type, metal layer thickness and design rule version identifier; the keyword matching unit matches the process feature words in the intermediate representation data through regular expression, associates the mapping relationship library of the process node and the analysis algorithm, generates the process node-analysis algorithm key-value pair and outputs to the strategy switching unit, realizes the accurate identification of the process feature and the dynamic mapping of the analysis strategy.

[0081] The strategy switching unit deploys the lightweight semantic analyzer defined by the dynamic adaptation module, loads the updated analysis algorithm and verification rule of the parameterized interface module. The lightweight semantic analyzer is based on a finite state machine model, and analyzes the context semantics of the process node-analysis algorithm key-value pair, and dynamically selects the analysis algorithm according to the priority of the process feature word; the verification rule is loaded from the rule library synchronized by the model self-optimization module, including the field separator fault tolerance range and the numerical accuracy verification threshold. After loading, the analysis algorithm and the verification rule are packaged as configuration parameter packages and output to the simulation conversion unit to drive the data analysis process of different processes.

[0082] The simulation conversion unit converts the parasitic parameters of different processes into a standardized simulation model through a process parameter abstraction layer. The process parameter abstraction layer defines the conversion rules of parasitic parameters for different process nodes, including resistance network topology mapping, capacitance coupling coefficient conversion, and dielectric loss model adaptation. During the conversion process, the metal layer parasitic capacitance is normalized to a standard unit based on the coupling coefficient, and the resistance network is reconstructed in topology structure according to the equivalent circuit model. The standardized simulation model is output to the visualization analysis module, which renders the abnormal area heat map and labels the process layer distribution. At the same time, through the adaptive interface of the parameterized interface module, the simulation model is converted into a netlist format compatible with the SPICE simulator, driving circuit simulation verification. Each unit cooperates through configuration parameter package and data flow to realize the end-to-end adaptive process of process feature recognition, analysis strategy switching, and simulation model generation.

[0083] In the specific implementation of the present application, the distributed sensor network is used to collect real-time capacitance parameter data sets and dynamically changing environmental variable data, and the network time protocol is used to synchronize the clocks of multiple source sensors, thereby eliminating time drift errors. After the original data is converted into digital signals by an analog-to-digital converter, a unified timestamp is added and stored as a time-series associated capacitance and environmental joint data set. The data preprocessing module parses the original file row by row through regular expressions and finite state machines, extracts capacitance values, resistance network topology, and process node attribute fields; the structure verification unit verifies the node connection logic and value range based on the YAML rule set dynamically loaded by the parameterized interface module, and the data cleaning unit performs unit uniform conversion and linear interpolation completion on abnormal fields to generate a standardized data table output to the fuzzy test engine module.

[0084] After receiving the standardized data, the fuzzy test engine module, the mutation case generation unit, based on the syntax tree mutation function, disturbs the numerical fields and topology connectors to generate a test case library including illegal characters, extreme values, and logical loops. The stress test unit injects the mutated cases into the analysis process of the parameterized interface module through an asynchronous data pipeline, monitors system crash points and timeout states, records error types and trigger conditions; the vulnerability mapping unit associates the test results with the case features, constructs a mapping relationship table of vulnerability types and trigger conditions, and pushes it to the abnormal response module. The parameterized interface module dynamically loads the analysis rule set according to the process node keyword, the intelligent identification unit matches the multi-way parser to convert the mixed delimiter, generates intermediate representation format data output to the abnormal response module and the visualization analysis module, and realizes the unified processing of multi-source heterogeneous data.

[0085] The abnormal response module embeds a probe detection unit in the analysis process, captures zero division errors and memory overflow operations in real time based on a vulnerability mapping table, modifies the decision unit to call an interpolation algorithm to complete repairable data and rewrite the analysis process, generates a report with error codes, and pushes it to the parasitic parameter extraction tool to terminate non-repairable tasks. The incremental update unit adds abnormal scene features to the model self-optimization module rule library, extracts high-frequency abnormal patterns based on DBSCAN clustering results, generates new analysis rules, and synchronizes them to the parameterized interface module to adjust the fault tolerance threshold. The dynamic adaptation module switches the analysis strategy according to the process keyword library, the lightweight semantic analyzer loads the updated algorithm and verification rules, the simulation conversion unit converts parasitic parameters of different processes into standardized simulation models through the process parameter abstraction layer, and outputs them to the visual analysis module to render a heat map and adapt to the SPICE simulator for circuit verification. The layout optimization suggestions generated by the visual analysis module are fed back to the model self-optimization module to drive the closed-loop iterative optimization process, continuously improving the analysis accuracy and robustness of the system in a dynamic environment.

[0086] The following describes the core algorithms, models and strategies involved in the present application:

[0087] The data preprocessing module uses regular expressions and finite state machines to work together. Regular expressions identify key fields such as capacitance values and resistance network topologies in the original file through predefined text patterns, acting as a filtering template to quickly locate target data. Finite state machines establish state transition logic for the analysis process, switching processing modes step by step when scanning data line by line, such as jumping from field recognition state to value verification state, forming a structured processing pipeline. This combination mechanism can effectively handle format differences in multi-source heterogeneous data.

[0088] The fuzzing engine module deploys a syntax tree mutation strategy. After standardizing the data into a tree-like logical structure, boundary perturbation operations are applied to numerical nodes, such as replacing normal capacitance values with extreme values beyond the process limits; logical destruction operations are implemented on topology connection nodes, such as deleting key connection points to create self-closing loops. By systematically constructing unconventional input combinations, a test case library covering illegal characters, value out-of-range and logical conflicts is generated, actively exposing potential system vulnerabilities.

[0089] The abnormal response module realizes abnormal capture based on the probe detection mechanism. Lightweight monitoring points are implanted in the analysis thread of the parameterized interface, and dangerous operation features are matched in real time according to the vulnerability mapping table pushed by the fuzzing engine. When detecting predefined risk behaviors such as zero division error or memory overflow, an interrupt protection process is triggered immediately. The repair decision unit compensates for repairable abnormalities using interpolation algorithms, such as filling in missing values based on adjacent node parameter trends; for non-repairable abnormalities, it generates a report with error trace information, forcing the termination of the current task to prevent error propagation.

[0090] The model self-optimization module uses a DBSCAN unsupervised clustering algorithm. The metal layer capacitance coupling strength and the abnormality occurrence frequency are taken as feature vectors, and a dense clustering is used to identify high-frequency fault areas. The algorithm automatically aggregates abnormal points with dense spatial distribution into fault clusters, such as a capacitance abnormal cluster caused by a specific metal layer thickness deviation, while filtering discrete accidental error points. The clustering results drive the rule updating unit to extract high-frequency fault patterns and generate targeted analysis rules.

[0091] The dynamic adaptation module uses a two-level strategy to achieve process compatibility. A lightweight semantic analyzer dynamically switches the analysis strategy according to process keywords, such as enabling a vertical line separator identification rule for a 5nm process, and using a comma separation processing logic for a 28nm process. The process parameter abstraction layer establishes a process conversion rule library, such as normalizing parasitic capacitance values of different processes through coupling coefficients, and reconstructing the topology of resistance networks according to equivalent circuit models, to form a standardized simulation model output.

[0092] The visual analysis module enhances the feedback mechanism through heat map mapping technology. Based on the force-directed algorithm to generate a circuit topology map, the abnormal area color rendering is superimposed on the layout coordinate system, with red highlighted areas representing high-frequency fault locations and blue transition areas indicating potential risk points. This spatial mapping method directly reveals the abnormal distribution of process parameters, providing spatial reference for layout optimization suggestions.

[0093] The closed-loop optimization mechanism runs through the system workflow. The incremental updating unit synchronizes the captured abnormal features to the rule library, driving the fuzzy test engine to dynamically adjust the mutation case weights; the model self-optimization module distributes new rules to the parameterized interface through the message queue; the dynamic adaptation module loads the updated analysis strategy and verifies the optimization effect through visual analysis, forming a continuous iterative technical enhancement loop. This design enables the system to have self-evolution ability to cope with unknown abnormal scenarios.

[0094] When implementing the present application in the field of integrated circuit design automation, the specific operation process is as follows: the original data file output by the parasitic parameter extraction tool is first input into the data preprocessing module. The syntax analysis unit uses a predefined regular expression pattern to identify capacitance values, resistance network topology, and process node attribute fields, and a finite state machine controls the state jump of the analysis process to achieve accurate field extraction. The structure verification unit loads the YAML rule set dynamically configured by the parameterized interface module, verifies the logical consistency of the node connection relationship, and checks whether the numerical range meets the definition of the current process node, and marks the abnormal fields that exceed the process limit value. The data cleaning unit performs unit standardization conversion on the verified data, unifies different orders of magnitude of capacitance values into picofarad units, and uses a linear interpolation algorithm to complete the missing data points based on the parameter trend of adjacent nodes, to form a standardized data table output to the fuzzy test engine module.

[0095] After receiving the standardized data, the fuzzing engine module constructs an abstract syntax tree structure. The numerical mutation subunit applies boundary perturbation to the capacitance and resistance value nodes, generating test cases containing negative values or exceeding the process limit value. The topology mutation subunit implements a logical destruction operation on the connection relationship node, creating abnormal topological structures such as node deletion or self-closing loops. The generated test case library is injected into the parameterized interface module through an asynchronous message queue to parse the process, and the stress test unit monitors the memory occupancy rate threshold and response timeout state in the process, recording error type codes. The vulnerability mapping unit associates error types with mutation case features, generating a vulnerability type and trigger condition mapping table and pushing it to the abnormal response module.

[0096] The rule loading unit in the parameterized interface module obtains the updated parsing rules from the model self-optimization module through the message queue. The intelligent recognition unit activates the corresponding parser instance based on the process keywords and converts the mixed delimiter data into a standard intermediate format. The format compatible unit processes SPICE netlists, CSV matrices, and custom binary streams in parallel, and generates unified intermediate representation data after deserialization. The abnormal response module implants a probe detection unit at the key nodes of the parsing thread, and matches zero division operations and memory overflow features in real time based on the vulnerability mapping table. The correction decision unit calls the interpolation algorithm to compensate for repairable anomalies, and generates a report with error trace information for non-repairable anomalies and terminates the task.

[0097] The topology graph rendering engine of the visual analysis module uses the force-directed algorithm to layout nodes, with node size mapping capacitance value intensity and edge width corresponding to resistance value size. In the layout coordinate system, a heat map layer is superimposed to mark the distribution density of abnormal areas with a color gradient. The clustering analysis unit of the model self-optimization module takes the spatial coordinates of abnormal points and the process layer number as feature vectors, and uses the DBSCAN algorithm to identify high-density fault clusters. The rule updating unit extracts the cluster center features to generate new parsing rules, and updates the fault tolerance threshold and delimiter definition of the parameterized interface module through the message queue.

[0098] The keyword matching unit of the dynamic adaptation module extracts process feature words from intermediate data and queries the process node-parsing algorithm mapping library. The strategy switching unit dynamically loads a lightweight semantic analyzer instance to parse the process conversion rules defined by the process parameter abstraction layer. The simulation conversion unit normalizes multi-process parasitic parameters into a standard simulation model, and outputs a SPICE-compatible netlist to drive circuit verification. The system continuously optimizes through an incremental update mechanism, with new features captured by the abnormal response module being synchronized to the rule library in real time, and the model self-optimization module dynamically adjusting clustering parameters, forming a closed-loop iterative technology evolution path.

[0099] The application solves the problem of insufficient robustness caused by the lack of processing capability of unexpected input scenarios in existing systems by constructing an automated abnormal data processing mechanism and a dynamic adaptation framework. The data preprocessing module uses regular expressions and finite state machines to parse the original data, extract the capacitance, resistance network topology and process node attribute fields, verify the node connection relationship and value range through the structure verification unit, and then the data cleaning unit performs unit uniform conversion and interpolation completion, and outputs standardized data to the fuzzy test engine module. The fuzzy test engine module generates a test case library including illegal characters, extreme values and logical loops through syntax tree mutation, the stress test unit injects the mutated cases into the parameterized interface module parsing process, monitors the system crash point and records the trigger condition, the vulnerability mapping unit constructs the mapping relationship table of vulnerability type and trigger condition, and pushes it to the abnormal response module to update the detection rules, forming a full coverage test and response mechanism for abnormal input scenarios.

[0100] The dynamic adaptation capability is realized through the cooperation of the parameterized interface module and the model self-optimization module. The parameterized interface module dynamically loads the YAML format parsing rule set based on the process node keyword, the rule loading unit extracts the field separator, numerical precision and topology connection symbol definition from the rule library synchronized by the model self-optimization module, the intelligent recognition unit matches the multi-way parser to convert mixed separators, and generates intermediate representation format data. The model self-optimization module uses unsupervised learning to cluster abnormal scenarios, extracts high-frequency abnormal patterns to generate new parsing rules, and distributes them to the parameterized interface module through the message queue to adjust the fault tolerance threshold. The dynamic adaptation module switches the parsing strategy according to the process keyword library, the lightweight semantic analyzer loads the updated parsing algorithm and verification rules, and the simulation conversion unit converts parasitic parameters of different processes into standardized simulation models through the process parameter abstraction layer, adapts to SPICE and FastHenry simulators, and realizes dynamic compatibility and accurate analysis of multiple process nodes.

[0101] The closed-loop iterative optimization mechanism drives the system to continuously optimize through the feedback link of the abnormal response module and the visualization analysis module. The probe detection unit embeds the probe to capture abnormal operations in the parsing and simulation stages, the correction decision unit calls the interpolation algorithm to complete the repairable data or generates an error report to terminate the task, and the incremental update unit synchronizes the abnormal scenario features to the rule library of the model self-optimization module. The visualization analysis module renders the abnormal area heat map and scatter plot, generates layout optimization suggestions and feeds them back to the model self-optimization module, triggers the clustering analysis unit to update the high-frequency abnormal pattern and the parsing rule. The updated rules and fault tolerance thresholds are synchronized to the parameterized interface module and the dynamic adaptation module through the message queue, forming a closed-loop link of testing, optimization and retesting, and improving the robustness and design iteration efficiency of the system in a dynamic environment.

Claims

1. A system for automated analysis of parasitic parameter results based on FUZZ testing, characterized in that, Comprise: Data preprocessing module, fuzzy testing engine module, parameterized interface module, abnormal response module, visual analysis module, model self-optimization module and dynamic adaptation module; The data preprocessing module is used for receiving the original file output by the parasitic parameter extraction tool, extracting the capacitance, resistance network topology and process node attribute field through the syntax analysis unit line by line scanning, verifying the node connection relationship and value range by the structure checking unit, and outputting the standardized data by the data cleaning unit after executing unit conversion and interpolation completion. The fuzzy testing engine module receives the standardized data, generates a test case library including illegal characters, extreme values and logical loops by the mutation case generation unit for syntax tree mutation of the numerical field and topology connection symbol, the stress test unit injects the mutation case into the analysis process and monitors the system crash point, the vulnerability mapping unit associates the error type and the trigger condition to generate a vulnerability mapping table, and pushes it to the abnormal response module; The parameterized interface module dynamically loads the YAML format analysis rule set based on the process node keyword, matches the multi-way parser conversion mixed delimiter through the intelligent recognition unit, generates intermediate representation format data and outputs it to the abnormal response module and the visual analysis module; The abnormal response module embeds a probe detection unit in the analysis and simulation stage according to the vulnerability mapping table to capture illegal operations, calls an interpolation algorithm to complete the repairable abnormal data in the correction decision unit, generates an exception report with error code, and pushes it to the parasitic parameter extraction tool to terminate the task, and the incremental update unit synchronizes the abnormal scene characteristics to the analysis rule library of the parameterized interface module; The visual analysis module receives the intermediate representation format data and the exception report, maps the parasitic parameter network through the topology graph rendering engine, superimposes the abnormal area heat map and scatter plot, and generates layout optimization suggestions to feed back to the model self-optimization module; The model self-optimization module uses unsupervised learning to cluster abnormal scenes, extracts high-frequency abnormal patterns to generate new analysis rules, distributes them to the parameterized interface module through the message queue to adjust the fault tolerance threshold, and synchronizes them to the dynamic adaptation module; The dynamic adaptation module matches the process feature words in the process keyword library, loads the corresponding analysis algorithm through the strategy switching unit, and converts different process parasitic parameters into standardized simulation models and adapts to SPICE simulation.

2. The FUZZ test based parasitic parameter result automated analysis system of claim 1, wherein, The data preprocessing module comprises: The syntax analysis unit receives the original file output by the parasitic parameter extraction tool, and extracts the capacitance, resistance network topology and process node attribute field through the syntax analysis function of the data preprocessing module by using regular expression and finite state machine line by line scanning, and outputs it to the structure checking unit; The structure checking unit verifies whether the node connection relationship and value range meet the process node definition based on the YAML format analysis rule set dynamically loaded by the parameterized interface module, generates a verification log to mark the abnormal field and passes it to the data cleaning unit; The data cleaning unit performs unit conversion and interpolation completion operation on the verified data output by the structure checking unit, generates a standardized data table and outputs it to the fuzzy testing engine module.

3. The FUZZ test based parasitic parameter result automated analysis system of claim 2, wherein, The fuzzy testing engine module comprises: The variation case generation unit receives the standardized data output by the data preprocessing module, disturbs the numerical fields and topological connectors based on the syntax tree variation function of the fuzzy test engine module, generates a test case library including illegal characters, extreme values and logical loops, and outputs to the stress test unit; the stress test unit injects the variation case into the parsing process of the parameterized interface module through the data pipeline defined by the fuzzy test engine module, monitors the system crash point and timeout state, records the error type and triggering condition and delivers to the vulnerability mapping unit; the vulnerability mapping unit associates the test results recorded by the stress test unit with the variation case features generated by the fuzzy test engine module, constructs a mapping relationship table of vulnerability types and triggering conditions, and pushes to the abnormal response module to update the detection rules.

4. The FUZZ test based parasitic parameter result automated analysis system of claim 3, wherein, The parameterized interface module comprises: The rule loading unit dynamically loads the field separator, numerical precision and topological connector definitions in YAML format based on the parsing rule library distributed by the model self-optimization module through the message queue, and outputs to the intelligent identification unit; The intelligent identification unit receives the test case library generated by the fuzzy test engine module, matches the multi-way parser defined by the parameterized interface module according to the process keywords, converts the mixed separator into intermediate representation format data, and outputs to the format compatible unit; The format compatible unit processes the SPICE netlist, CSV matrix and custom binary format data in parallel, generates unified intermediate representation format data and outputs to the abnormal response module and the visual analysis module.

5. The FUZZ test based parasitic parameter result automated analysis system of claim 4, wherein, The abnormal response module comprises: a probe detection unit, which embeds a probe in the parsing process and simulation stage of the parameterized interface module, captures division by zero errors and memory overflow operations in real time based on the vulnerability mapping table generated by the fuzzy test engine module, and delivers the detection results to the correction decision unit; the correction decision unit calls the interpolation algorithm of the data preprocessing module to complete the data for repairable anomalies, and writes back to the parsing process of the parameterized interface module for re-parsing, generates an error code report for unrepairable anomalies and pushes it to the parasitic parameter extraction tool to terminate the task; the incremental update unit adds the abnormal scenario features captured by the probe detection unit to the rule library of the model self-optimization module, adjusts the variation case generation weight of the fuzzy test engine module based on the clustering analysis result of the model self-optimization module, and synchronizes to the parameterized interface module through the message queue.

6. The FUZZ test based parasitic parameter result automated analysis system of claim 5, wherein, The model self-optimization module comprises: a clustering analysis unit that receives abnormal scenario data detected by the abnormal response module, generates layout optimization suggestions based on the visual analysis module, and clusters metal layer capacitance coupling features and high-frequency abnormal patterns using the DBSCAN algorithm in unsupervised learning, and outputs the clustering result to the rule update unit; The rule update unit generates new parsing rules according to the clustering result of the clustering analysis unit, adjusts the fault tolerance threshold based on the vulnerability mapping table of the fuzzy test engine module, and distributes to the parameterized interface module and the fuzzy test engine module through the message queue; The feedback closed loop unit synchronizes the updated parsing rules and fault tolerance threshold to the dynamic adaptation module through the message queue of the model self-optimization module, and triggers the visual analysis module to re-render the abnormal area heat map.

7. The FUZZ test based parasitic parameter result automated analysis system of claim 6, wherein, The dynamic adaptation module comprises: A keyword matching unit receives the uniform intermediate representation format data generated by the parameterized interface module, extracts process feature words based on the process keyword library synchronized by the model self-optimization module, matches the mapping library of process nodes and analysis algorithms, and outputs to a strategy switching unit; The strategy switching unit deploys the lightweight semantic analyzer defined by the dynamic adaptation module, dynamically loads the updated analysis algorithms and verification rules of the parameterized interface module, and outputs to a simulation conversion unit; The simulation conversion unit converts parasitic parameters of different processes into standardized simulation models through the process parameter abstraction layer of the dynamic adaptation module, outputs to a visual analysis module to render an abnormal area heat map, and adapts SPICE simulation through the parameterized interface module.

Images