Design consultation-oriented knowledge reasoning method and system

Abstract

The invention relates to the technical field of design consultation, and discloses a design consultation-oriented knowledge reasoning method, which comprises the following steps of: obtaining diversified knowledge data and preprocessing to generate a reference knowledge base; extracting knowledge association features to determine inference rules; constructing a reasoning chain and reasoning nodes; verifying effectiveness of the reasoning nodes, and distinguishing normal and abnormal reasoning states; reasoning according to a hierarchical sequence in a normal state, identifying abnormal nodes in an abnormal state, calculating a reasoning deviation value, and performing path correction until reasoning is completed. The system comprises a data acquisition module, a rule output module, a reasoning generation module, a state evaluation module and a reasoning optimization module. By integrating diversified data, scientifically formulating inference rules, dynamically optimizing inference paths and accurately processing exceptions, the efficiency and accuracy of design consultation knowledge inference are improved, and the method is suitable for intelligent inference decision in the field of design consultation.

Description

Technical Field

[0001] The present invention relates to the technical field of design consulting, and in particular to a knowledge reasoning method and system oriented to design consulting. Background Art

[0002] In the field of design consulting, with the rapid development of information technology and the increasing complexity of industry demands, how to efficiently integrate and utilize diverse knowledge data to achieve accurate and intelligent reasoning and decision-making has become a key issue that needs to be addressed. Traditional knowledge reasoning methods have exposed many limitations when dealing with the complex knowledge systems in design consulting.

[0003] From the perspective of knowledge data acquisition and processing, design consulting needs typically involve multiple types of knowledge data from multiple sources, including design requirement documents, historical case data, and industry specifications. However, existing methods struggle to systematically integrate this diverse data, plagued by significant data redundancy and conflicts, and lack effective semantic standardization. This leads to a chaotic structure of knowledge units, making it impossible to form a unified, standardized baseline knowledge base, severely impacting the accuracy and efficiency of subsequent reasoning.

[0004] When it comes to extracting knowledge-related features and determining inference rules, traditional methods often ignore the differences in semantic relevance and strength between knowledge nodes, making it impossible to scientifically classify and prioritize knowledge-related features. For example, the failure to clearly distinguish the inference priorities of aggregated and dispersed features results in a lack of hierarchy and logic in the inference process, making it difficult to formulate reasonable inference rules based on the inherent relevance of knowledge. This leads to irrational inference path construction and low inference efficiency.

[0005] When it comes to reasoning path construction and reasoning node management, existing technologies lack a dynamic adjustment mechanism when constructing reasoning chains and reasoning nodes. When abnormal reasoning nodes appear during the reasoning process, they cannot be identified and processed promptly and accurately. For example, the reasoning deviation cannot be effectively calculated, and it is difficult to correct the path based on the deviation, causing the reasoning process to easily become deadlocked and unable to guarantee the consistency and reliability of the reasoning. At the same time, the means of verifying the validity of the reasoning nodes are single, and there is a lack of scientific verification models and parameter evaluation systems. This makes it difficult to accurately judge the status of the reasoning nodes in real time, and it is impossible to detect abnormal situations in the reasoning process in a timely manner.

[0006] Furthermore, traditional methods lack path correction strategies based on historical data and dynamic adjustment factors when handling abnormal reasoning nodes. This inability to conduct comprehensive analysis and correction based on multi-dimensional data, such as the number of historical backtracking times and reasoning deviations at abnormal reasoning nodes, results in insufficiently targeted and effective exception handling, making it difficult to quickly restore the normal reasoning process, impacting the efficiency and quality of the entire design consulting knowledge reasoning process. Summary of the Invention

[0007] The purpose of the present invention is to provide a knowledge reasoning method and system for design consultation to solve the problems raised in the above background technology.

[0008] To achieve the above-mentioned object, the present invention provides the following technical solution: a knowledge reasoning method for design consultation, the method comprising: Acquire diversified knowledge data for design consulting needs, perform knowledge preprocessing, and generate a benchmark knowledge base; Extracting knowledge association features of each of the benchmark knowledge bases, and determining inference rules for different knowledge bases based on the knowledge association features; Constructing reasoning paths for each benchmark knowledge base according to the reasoning rules to form multiple groups of reasoning chains and reasoning nodes for each of the reasoning chains; Performing validity verification on each of the inference nodes to obtain an inference state of each node, wherein the inference state includes a normal inference state and an abnormal inference state; In the normal reasoning state, the reference knowledge bases under each reasoning node perform associative reasoning in a hierarchical order; In the abnormal reasoning state, identifying the abnormal reasoning node and calculating the reasoning deviation of the benchmark knowledge base under the abnormal reasoning node; Based on the inference deviation, the path of the abnormal inference node is corrected until all the benchmark knowledge bases complete the inference in hierarchical order and terminate.

[0009] Preferably, the acquisition of diversified knowledge data of design consulting needs and knowledge preprocessing to generate a benchmark knowledge base includes: Collect design requirement documents, historical case data, and industry specification data, integrate them into independent knowledge units, and assign a unique identifier to each of the knowledge units; Performing redundancy filtering on the data within each of the independent knowledge units to remove duplicate or conflicting knowledge entries; The filtered independent knowledge units are semantically standardized to unify the data structures of the knowledge units, and the structurally unified independent knowledge units are output as a benchmark knowledge base.

[0010] Preferably, the extracting of knowledge association features of each of the benchmark knowledge bases includes: Obtaining the knowledge nodes under each of the benchmark knowledge bases and sorting them according to the strength of association; Extracting semantic relevance between adjacent knowledge nodes and marking it as classification condition parameter; Obtaining a preset correlation threshold, dividing each classification condition parameter into intervals according to the correlation threshold, and generating multiple correlation intervals; Counting the number of classification condition parameters within each of the association intervals and recording them as feature distribution parameters, and determining the knowledge association features of each benchmark knowledge base based on the feature distribution parameters; The knowledge association features include aggregation features and dispersion features, and the reasoning priority of the dispersion features is lower than the reasoning priority of the aggregation features.

[0011] Preferably, the determining of the knowledge association features of each benchmark knowledge base based on the feature distribution parameters includes: Obtaining feature distribution parameters under each of the benchmark knowledge bases; Sort the feature distribution parameters under the same benchmark knowledge base by value from high to low, and calculate the weight of the feature distribution parameter with the highest value; Obtaining a weight threshold, and comparing the weight threshold with the weight of the characteristic distribution parameter with the highest value; If the proportion weight is greater than the weight threshold, it is determined that the reference knowledge base corresponding to the feature distribution parameter has an aggregation feature; If the proportion weight is less than or equal to the weight threshold, it is determined that the reference knowledge base corresponding to the feature distribution parameter has a dispersion feature.

[0012] Preferably, the determining of inference rules of different knowledge bases based on the knowledge association features includes: Acquiring knowledge association features of each of the benchmark knowledge bases; The maximum value of the correlation interval corresponding to the aggregation feature is used as the hierarchical reasoning interval; Summarize the hierarchical reasoning intervals of the benchmark knowledge base under all the aggregation features, and arrange them from small to large according to the interval values to generate a reasoning priority sequence; Performing hierarchical reasoning on each of the benchmark knowledge bases according to the reasoning priority sequence, and after the reasoning of the benchmark knowledge base under the aggregation feature is completed, additional reasoning is performed on the benchmark knowledge base under the dispersion feature; Generating the reasoning priority sequence includes: Obtaining historical adjustment records of hierarchical reasoning intervals of a benchmark knowledge base under the aggregation feature; Extract the actual effective duration of the hierarchical reasoning interval in the historical reasoning path and calculate the deviation ratio between it and the preset duration; Generating a dynamic adjustment factor based on the deviation ratio, and performing a weighted correction on the current hierarchical reasoning interval according to the adjustment factor; Summarize the weighted hierarchical reasoning intervals, rearrange the reasoning priority sequence, and synchronize the adjusted sequence to the path construction of subsequent reasoning nodes.

[0013] Preferably, the performing validity verification on each of the inference nodes includes: Obtaining the inference results of the benchmark knowledge base under each of the inference nodes, and performing vector mapping conversion to generate multiple verification vectors; Calling a preset verification function, inputting the verification vector into the verification function, and recording the output value as a verification parameter; Obtaining a verification threshold, and comparing the verification parameter with the verification threshold; If the verification parameter is less than the verification threshold, it is determined that the inference node is in a normal inference state; If the verification parameter is greater than or equal to the verification threshold, it is determined that the inference node is in an abnormal inference state.

[0014] Preferably, the calculating of the reasoning deviation of the benchmark knowledge base at the abnormal reasoning node includes: Obtaining an actual reasoning node of a benchmark knowledge base for which associated reasoning has not been completed under the abnormal reasoning node; Calculating the logical interval between the actual reasoning node and the abnormal reasoning node, and calibrating it as a real-time deviation; Perform historical backtracking on the abnormal reasoning node and extract the historical deviation of its associated historical nodes; Obtaining a preset fault tolerance threshold, and terminating the backtracking operation when the historical deviation is less than or equal to the fault tolerance threshold; The deviation measurement function is called, the real-time deviation amount and the historical deviation amount are input into the measurement function, and the output result is recorded as the inferred deviation amount.

[0015] Preferably, the performing path correction on the abnormal reasoning node based on the reasoning deviation includes: Count the historical backtracking times of the abnormal reasoning node and record it as an adjustment reference value; Obtaining an adjustment threshold, and comparing the adjustment reference value with the adjustment threshold; When the adjustment reference value is greater than or equal to the adjustment threshold, performing a logical shift on the abnormal reasoning node according to the reasoning deviation amount, generating a corrected reasoning node, and performing associated reasoning under the corrected reasoning node; When the adjustment reference value is less than the adjustment threshold, the inference deviation is continuously collected until the adjustment reference value reaches or exceeds the adjustment threshold, and then the path correction is triggered.

[0016] Preferably, the vector mapping conversion includes: Extract semantic entities and relationship attributes from the reasoning results and construct multidimensional feature vectors; Performing dimensionality reduction processing on the multidimensional feature vector to generate a standardized mapping result in a low-dimensional vector space; The standardized mapping result is weightedly calculated according to a preset semantic weight matrix to generate the verification vector and input it into the verification function.

[0017] Preferably, the present invention further includes a knowledge reasoning system for design consultation, the system comprising: A data acquisition module, which is used to acquire diversified knowledge data of design consulting needs, perform knowledge preprocessing, and generate a benchmark knowledge base; A rule output module, the rule output module is used to extract the knowledge association features of each of the benchmark knowledge bases and determine the inference rules of different knowledge bases based on the knowledge association features; An inference generation module, configured to construct an inference path for each benchmark knowledge base according to the inference rules, forming multiple groups of inference chains and inference nodes for each of the inference chains; A state evaluation module, the state evaluation module is used to verify the validity of each of the inference nodes to obtain the inference state of each node; In the normal reasoning state, the reference knowledge bases under each reasoning node perform associative reasoning in a hierarchical order; In the abnormal reasoning state, identifying the abnormal reasoning node and calculating the reasoning deviation of the benchmark knowledge base under the abnormal reasoning node; The inference optimization module is used to correct the path of the abnormal inference node based on the inference deviation until all the benchmark knowledge bases complete the inference in hierarchical order and terminate.

[0018] Compared with the prior art, the present invention has the following beneficial effects: In the knowledge data processing phase, design requirements documents, historical case data, and industry standard data are collected, integrated into independent knowledge units, and assigned unique identifiers, effectively achieving systematic management of diverse knowledge data. Redundancy filtering removes duplicate or conflicting knowledge entries, ensuring the accuracy and consistency of knowledge data. Semantic standardization unifies the data structure of knowledge units, generating a standardized and orderly baseline knowledge base. This lays a solid foundation for subsequent reasoning, avoids reasoning errors caused by data confusion, and improves the usability of knowledge data and the reliability of reasoning starting points.

[0019] In terms of knowledge association feature extraction and reasoning rule determination, by obtaining knowledge nodes and sorting them by association strength, extracting semantic association and dividing them into intervals, the aggregation and dispersion features of the benchmark knowledge base can be accurately identified. It is clear that the reasoning priority of dispersion features is lower than that of aggregation features, making the reasoning process more hierarchical and logical. The knowledge association features are determined based on feature distribution parameters and weight thresholds to ensure the scientific nature of the reasoning rule formulation. The introduction of historical adjustment records and the actual effective time deviation ratio to generate dynamic adjustment factors, weighted correction of hierarchical reasoning intervals, realizes the dynamic optimization of the reasoning priority sequence, enables the reasoning rules to be flexibly adjusted according to the actual reasoning situation, and improves the adaptability of the reasoning rules and the rationality of the reasoning process.

[0020] During the inference path construction process, multiple inference chains and inference nodes are formed according to inference rules, establishing a clear inference context. By verifying the validity of inference nodes, using vector mapping conversion, preset verification functions, and verification thresholds, the normal or abnormal status of inference nodes can be accurately determined in real time. This provides strong support for timely detection of problems in the inference process, ensuring the validity of the inference path and the controllability of the inference process.

[0021] In handling abnormal inference nodes, the real-time deviation is calculated by calculating the logical interval between the actual inference node and the abnormal inference node. Combined with the historical deviation extracted through historical backtracking, the deviation measurement function accurately calculates the inference deviation, providing a quantitative basis for exception handling. Based on the comparison of the number of historical backtracking times with the adjustment threshold, different path correction strategies are implemented to achieve precise correction of abnormal inference nodes. Logical offsets generate corrected inference nodes and execute associated inferences, or continuously collect inference deviations until correction is triggered. This ensures the targeted and effective handling of exceptions, improves the fault tolerance and recovery efficiency of the inference process, and ensures the consistency and reliability of reasoning. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a working principle diagram of the knowledge reasoning method for design consultation according to the present invention; Figure 2 Design graphs generated for the benchmark knowledge base; Figure 3 Design diagram for determining knowledge association features; Figure 4 Design diagram for validating the inference node; Figure 5 Design diagram for abnormal reasoning node path correction. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] See also Figure 1-Figure 5 The present invention relates to a knowledge reasoning method for design consulting. The technical solution is described in detail below in conjunction with specific implementation methods.

[0025] The system collects diverse knowledge data from design consulting scenarios through interfaces, including design requirement documents, historical case data, and industry specification data. First, each type of data is integrated into independent knowledge units, and each knowledge unit is assigned a unique identifier to enable precise positioning and management. Redundancy filtering is then performed on the data within the independent knowledge units, and duplicate or conflicting knowledge entries are removed through a data comparison algorithm. For example, different expressions of the same parameter in the design specification are calibrated using an industry standard library. After filtering is complete, the knowledge units are semantically standardized and the data structure is unified. For example, design parameters in different formats are converted to the standard format defined by XMLSchema, and the final output is a unified structural benchmark knowledge base.

[0026] The system parses each benchmark knowledge base, extracts the knowledge nodes therein (such as design parameters, case features, specification clauses, etc.), and calculates the correlation strength between the nodes based on the cosine similarity algorithm, sorting them from high to low by strength value. For adjacent knowledge nodes, the semantic correlation is calculated using a semantic analysis model (such as Word2Vec) to generate classification condition parameters. A preset correlation threshold (such as 0.6) is used to divide the classification condition parameters into intervals, for example, into a high correlation interval (≥0.8), a medium correlation interval (0.6-0.8), and a low correlation interval (<0.6). The number of parameters in each interval is counted to form a feature distribution parameter (such as the high correlation interval parameters account for 40%), based on which the knowledge correlation feature is judged to be aggregated or dispersed (see the subsequent embodiments for specific judgment rules). The inference rule is determined according to the feature type: the knowledge base corresponding to the aggregate feature adopts the hierarchical inference rule, and the knowledge base corresponding to the dispersed feature adopts the supplementary inference rule.

[0027] For the benchmark knowledge base with aggregated features, the maximum value of the correlation interval is used as the hierarchical reasoning interval (for example, if the maximum value of the high correlation interval is 0.9, the interval is 0.9), and the reasoning priority sequence is generated from small to large according to the interval value. For example, the hierarchical reasoning intervals of three aggregated knowledge bases are 0.5, 0.7, and 0.9, respectively, and the priority sequence is knowledge base A (0.5) → knowledge base B (0.7) → knowledge base C (0.9). The reasoning chain is constructed according to the priority order, and each reasoning step corresponds to an inference node. The node contains the knowledge unit and association rules currently being reasoned. For knowledge bases with decentralized features, after the inference of the aggregated knowledge base is completed, additional reasoning is performed in parallel to form a branching reasoning chain.

[0028] The system obtains the output results of the inference nodes, extracts semantic entities and relationship attributes, and constructs a multidimensional feature vector (for example, a design parameter vector includes dimensions such as material, size, and performance). Principal component analysis (PCA) is used to reduce the dimensionality of the vector and map it to a low-dimensional vector space. A weighted calculation is then performed using a preset semantic weight matrix (e.g., 30% weight for design specifications and 50% weight for case experience) to generate a verification vector. A preset verification function (e.g., an SVM classifier) is called to evaluate the verification vector and output verification parameters (e.g., classification probability values). A preset verification threshold (e.g., 0.7) is used. If the verification parameter is <0.7, the inference state is considered normal; if it is ≥0.7, the inference state is considered abnormal.

[0029] During normal reasoning, the benchmark knowledge bases under each inference node perform associated reasoning sequentially according to the inference priority sequence and hierarchical reasoning interval. For example, after knowledge base A in the priority sequence completes reasoning for the current node, it triggers reasoning for knowledge base B after an interval of 0.5 time units (dynamically adjusted based on computing resources), and so on, forming a linear and progressive reasoning chain.

[0030] When an abnormal reasoning node is detected, the system first locates the benchmark knowledge base for the incomplete associated reasoning, and calculates the logical interval between the actual reasoning node and the abnormal node (such as the difference in reasoning steps) as the real-time deviation. At the same time, the abnormal node is historically backtracked, and the historical deviation of the associated historical nodes is extracted. The backtracking is stopped when the historical deviation is ≤ the preset fault tolerance threshold (such as 2 steps). The deviation measurement function (such as the mean square error function) is called to combine the real-time deviation and the historical deviation to obtain the reasoning deviation. Path correction is performed based on the deviation: if the number of historical backtracking times is ≥ the adjustment threshold (such as 3 times), the abnormal node is logically offset, a corrected reasoning node is generated (such as skipping the current conflicting rule and enabling the backup rule), and reasoning is continued under the corrected node; if the number of backtracking times is less than the adjustment threshold, the deviation is continuously collected until the threshold is reached and correction is triggered.

[0031] The present invention will be further described below in conjunction with Examples 1 to 5: Example 1: In the process of acquiring diversified knowledge data of design consulting needs and performing knowledge preprocessing to generate a benchmark knowledge base, the system performs the following operations: To capture diverse knowledge data, the system automatically collects design requirements documents, historical case data, and industry specification data through multi-source data interfaces. Design requirements documents contain text-based user requirement descriptions, such as product function definitions, performance requirements, and design constraints. The system parses the document content using natural language processing (NLP) technology and utilizes regular expressions to extract key parameters and entity information. For example, from the query "need to design an industrial shelf with a load capacity ≥ 500kg," the system extracts key elements such as "load capacity" and "500kg." Historical case data is sourced from an internal case database and contains structured fields such as case number, project background, design process records, and final solution parameters. The system directly accesses this data through database connection protocols (such as JDBC) to ensure the integrity and consistency of case data. Industry specification data is connected to external authoritative standards platforms (such as the National Standards Full Text Public Disclosure System) and synchronized with specification terms in real time through API interfaces. These include design standards, safety regulations, and material selection rules, such as fire protection regulations in architectural design and tolerance and coordination standards in mechanical design.

[0032] When integrating various types of data into independent knowledge units, the system assigns a unique identifier to each knowledge unit. The identifiers use a hierarchical coding structure, such as "D-20250528-001-01," where "D" represents the design requirement category, "20250528" is the data collection date, "001" is the sequence number within the general category, and "01" is the sub-item number of the knowledge unit within the general category. This coding rule facilitates the subsequent rapid location, retrieval, and management of knowledge units, ensuring that data from different sources has unique identities within the benchmark knowledge base.

[0033] After completing the initial integration of knowledge units, the system performs redundant filtering operations on the data within each independent knowledge unit. Redundant filtering is divided into two stages: the first stage is vertical filtering, which uses a hash value comparison algorithm to detect duplicate records for data within the same knowledge unit. For example, there may be multiple identical performance indicator descriptions in a design requirements document. The system calculates the hash value of each record and determines that records with exactly the same hash value are duplicates and deletes them; the second stage is horizontal filtering, which cross-validates data from different knowledge units to identify conflicting entries. For example, in a historical case, a design parameter value is "10mm", while the industry specification requires that the parameter value range is "15-20mm". At this time, the system marks the conflicting entry as pending and triggers a manual review process. The design consultant confirms and deletes the erroneous entry or adjusts the parameter value.

[0034] During the semantic standardization phase, the system unifies the data structure of filtered independent knowledge units based on a predefined knowledge meta-model. This knowledge meta-model employs the entity-attribute-value (EAV) model architecture, abstracting various types of knowledge into entity objects. Each entity contains several attributes, each corresponding to a specific value. For example, "steel" is defined as an entity, with attributes such as "grade," "yield strength," and "elongation." Attribute values are standardized based on the data source. For unstructured text data in design requirements documents, named entity recognition (NER) is used to extract entities and attributes. For example, from the sentence "Select Q235B steel, with a yield strength not less than 235 MPa," the entity "Q235B steel," the attribute "yield strength," and its value "≥235 MPa" are identified and converted into the standard format of the EAV model. For structured data in historical cases, data mapping rules are used to map existing fields to corresponding attributes in the knowledge meta-model. For example, the "design cycle days" field in a case study is mapped to the "time dimension - cycle" attribute.

[0035] During the semantic standardization process, the system also needs to address terminology discrepancies across different data sources. For example, design requirements might use "floor area ratio," while industry standards describe it as "gross building area density." The system establishes a term mapping table to associate synonyms or near-synonyms with unified standard terms, ensuring semantic consistency across different knowledge units. The term mapping table utilizes a bidirectional mapping structure, supporting both forward queries (from business terms to standard terms) and reverse queries (from standard terms to business terms), facilitating semantic conversion and association during reasoning.

[0036] After completing the above processing, the system outputs the structurally unified independent knowledge units as a baseline knowledge base. The baseline knowledge base is stored in a graph database (such as Neo4j), with knowledge units as nodes in the graph. Semantic relationships (such as "belongs to," "associated with," and "constrained") connect nodes to form a knowledge network. For example, design requirement nodes and industry specification nodes are connected by a "conforms to" relationship, and historical case nodes and design parameter nodes are connected by a "contains" relationship. This storage structure facilitates the subsequent extraction of knowledge association features and the construction of reasoning paths.

[0037] During data collection and preprocessing, the system also features data quality monitoring. By setting monitoring indicators for data integrity, consistency, and accuracy, it can detect anomalies in the data collection and processing process in real time. For example, if the missing value rate for a key attribute of a knowledge unit exceeds a preset threshold (e.g., 20%), the system automatically triggers a supplementary data collection process, prompting relevant personnel to supplement the data via email or push notifications. If the semantically standardized data structure of a knowledge unit is detected to be inconsistent with the knowledge metamodel definition, the system generates an error log and locates the problem node for technical personnel to troubleshoot and repair.

[0038] In addition, the system supports version management for the baseline knowledge base. Each time the knowledge base is updated (e.g., adding new knowledge units, modifying attribute values, deleting conflicting entries, etc.), a new version number is automatically generated in the format of "major version number.minor version number.revision number" (e.g., V1.0.1). The update time, content, and operator information are also recorded. This version management feature ensures that the evolution of knowledge can be traced during reasoning. It also supports rollback to historical versions, facilitating the handling of reasoning anomalies caused by data updates.

[0039] Example 2: In the process of extracting knowledge association features from each benchmark knowledge base, the system analyzes the association strength of knowledge nodes and determines the feature type through the following steps: The system performs structured analysis on each benchmark knowledge base and identifies the knowledge nodes therein. The type of knowledge node is determined by the content attributes of the knowledge base. For example, in the benchmark knowledge base of architectural design consulting, knowledge nodes may include specific concepts or parameters such as "fire protection distance", "sunlight standard", "concrete strength grade", and "unit area". For design requirement knowledge bases, nodes may involve functional requirements proposed by users (such as "seismic fortification intensity" and "intelligent system integration requirements"); for historical case knowledge bases, nodes include key design elements in the case (such as "structural form", "material selection", and "project cycle"); nodes in industry specification knowledge bases are usually core indicators in the specification clauses (such as "load value", "safety factor", and "environmental protection requirements").

[0040] The system calculates the strength of association between each knowledge node and ranks them. The calculation of association strength is based on a semantic analysis model, using Word2Vec word vector technology to map each knowledge node into a vector in a high-dimensional space. The cosine similarity formula is used to calculate the cosine value of the angle between any two node vectors, which is used as a quantitative indicator of association strength. For example, the vector of node A's "fire separation distance" is [0.2, 0.5, 0.3, ...], and the vector of node B's "building height" is [0.3, 0.4, 0.3, ...]. The cosine similarity between the two is 0.82, indicating a strong association; while the cosine similarity between node C's "unit area" and node D's "steel yield strength" is 0.35, indicating a weak association. After calculating the association strength of all node pairs, the system generates a list of node association strengths from high to low according to the cosine similarity value, clearly sorting the closeness between nodes.

[0041] When extracting the semantic relevance between adjacent knowledge nodes, the system uses the association strength list as a basis to screen out adjacent node pairs (i.e., nodes directly preceding and following the sorted nodes) and further analyze their semantic relevance properties. The calculation of semantic relevance is integrated with the domain ontology model, utilizing the conceptual relationships defined in the ontology (such as "subclass relationship," "attribute relationship," and "constraint relationship") to semantically annotate node pairs. For example, in the building code ontology, "fire separation distance" and "building height" are related through a "positive correlation" relationship, with a quantifiable semantic relevance of 0.8; while "unit area" and "room layout" are related through a "containment relationship," with a semantic relevance of 0.65. These semantic relevance values serve as classification condition parameters for subsequent feature analysis.

[0042] The system presets a correlation threshold (e.g., 0.6) to divide the classification condition parameters into intervals. The setting of the correlation threshold is based on the experience of domain knowledge experts and can be dynamically adjusted according to different industries or design consulting types. With 0.6 as the dividing point, the classification condition parameters are divided into a high correlation interval (≥0.8), a medium correlation interval (0.6-0.8), and a low correlation interval (<0.6). For example, if the semantic correlation values of adjacent node pairs in a benchmark knowledge base are 0.85, 0.72, 0.58, and 0.35, respectively, they are classified as high correlation interval, medium correlation interval, low correlation interval, and low correlation interval, respectively.

[0043] The system counts the number of classification condition parameters within each correlation interval and generates feature distribution parameters. These parameters are presented as the percentage of parameters in each interval relative to the total number of parameters. For example, in the example above, the high correlation interval accounts for 25% (1 / 4), the medium correlation interval accounts for 25% (1 / 4), and the low correlation interval accounts for 50% (2 / 4). These parameters reflect the distribution characteristics of the degree of correlation among knowledge nodes in the benchmark knowledge base and are key to determining the type of knowledge correlation feature.

[0044] When determining the type of knowledge association feature, the system first obtains the feature distribution parameters of each benchmark knowledge base, sorts the feature distribution parameters under the same knowledge base from high to low by value, and calculates the weight of the feature distribution parameter with the highest value. For example, the feature distribution parameters of a knowledge base are 40% for the high-association interval, 30% for the medium-association interval, and 30% for the low-association interval. After sorting, the highest value is 40% of the high-association interval, and its weight is 40%. The system presets a weight threshold (such as 50%) and compares the weight of the highest value with the weight threshold: if the weight is greater than the weight threshold, the benchmark knowledge base is judged to have an aggregation feature, indicating that its knowledge nodes are mainly concentrated in the high-association or medium-high-association intervals, the knowledge structure is compact, and it is convenient for hierarchical reasoning; if the weight is less than or equal to the weight threshold, it is judged to be a dispersed feature, indicating that the knowledge nodes are relatively dispersed and the degree of association is low, and supplementary reasoning rules need to be adopted.

[0045] For example, consider a benchmark knowledge base for architectural design consulting. Assume the base contains 100 pairs of adjacent knowledge nodes, with 65 (65%) having high-correlation interval parameters, 20 (20%) having medium-correlation interval parameters, and 15 (15%) having low-correlation interval parameters. The characteristic distribution parameters are ranked numerically as 65%, 20%, and 15%, with the highest value being 65% and the weight being 65%, exceeding the preset weight threshold of 50%. Therefore, the knowledge base is considered to have a clustering characteristic. Conversely, if a mechanical design knowledge base has a high-correlation interval parameter ratio of 40%, a medium-correlation interval ratio of 35%, and a low-correlation interval ratio of 25%, and the highest value of 40% is less than the weight threshold of 50%, then the knowledge base is considered to have a dispersed characteristic.

[0046] When determining reasoning priority, the benchmark knowledge base with aggregated features is given a higher reasoning priority due to its tightly linked knowledge nodes and clear reasoning paths. The system uses the maximum value of the correlation interval corresponding to the aggregated features as the hierarchical reasoning interval. For example, the maximum value of the high correlation interval for the aforementioned architectural design knowledge base is 0.85, and the hierarchical reasoning interval is set to 0.85 units (units can be defined as milliseconds, seconds, etc. based on the system's calculation cycle). Knowledge bases with dispersed features have a lower reasoning priority than knowledge bases with aggregated features due to their loose node connections and high uncertainty in reasoning paths. They are usually initiated as a supplementary link after aggregated reasoning is completed to provide multi-angle reasoning support or verification.

[0047] Furthermore, the system supports dynamic updates of knowledge association features. When the baseline knowledge base content changes (such as the addition of new knowledge units or modification of node attributes), it automatically triggers the recalculation of association strength and semantic relevance, updates feature distribution parameters and feature type determination results, and ensures that knowledge inference rules are always based on the latest knowledge base structure. This dynamic mechanism enables the system to adapt to the rapid update and evolution of knowledge in the design consulting field, maintaining the accuracy and effectiveness of the inference process.

[0048] Example 3: When determining the inference rules of different knowledge bases based on knowledge association characteristics, the system dynamically generates and optimizes the inference rules through the following process: The system extracts the knowledge association features of each benchmark knowledge base, distinguishing between aggregated and dispersed features. For benchmark knowledge bases with aggregated features, their knowledge nodes are tightly linked, making hierarchical reasoning suitable. This involves performing reasoning sequentially based on the strength of the knowledge node associations. For knowledge bases with dispersed features, due to their looser node connections, they are typically used as supplementary reasoning steps, initiated after aggregated reasoning is complete.

[0049] For the benchmark knowledge base with aggregation features, the system uses the maximum value of the corresponding correlation interval as the hierarchical reasoning interval. The maximum value of the correlation interval reflects the strongest correlation degree of the knowledge nodes in the knowledge base. For example, the maximum value of the high correlation interval of a certain aggregation knowledge base is , then the hierarchical reasoning interval is set to ,in is the time unit conversion factor (for example, when converting the correlation value to seconds, milliseconds / unit correlation). This setting makes the inference interval positively correlated with the correlation strength of knowledge nodes. The stronger the correlation, the shorter the inference interval between nodes, and the more compact the inference process.

[0050] Next, the system summarizes the hierarchical reasoning intervals of the benchmark knowledge base under all aggregated features ( For the knowledge base index, ), and arrange them from small to large according to the interval value to generate the initial reasoning priority sequence (in ). For example, the hierarchical reasoning intervals of the three aggregated knowledge bases are Second, Second, seconds, the initial sequence is , the corresponding reasoning order is knowledge base A→knowledge base B→knowledge base C.

[0051] In order to optimize the reasoning priority sequence, the system introduces a dynamic adjustment mechanism for historical data. First, the hierarchical reasoning interval historical adjustment records of the benchmark knowledge base under the aggregation feature are obtained. Each record contains a timestamp, the original interval value, , adjusted interval value and actual effective time Then, extract the actual effective duration of the hierarchical reasoning interval in the historical reasoning path , and calculate its difference with the preset duration The deviation ratio , the calculation formula is: ; in, represents the deviation ratio (dimensionless), is the actual effective duration (unit: seconds), For the preset duration (unit: seconds), take the initial inference interval As The default value of .

[0052] Based on the deviation ratio , the system generates a dynamic adjustment factor The adjustment factor is calculated using a linear mapping method, for example, hour (No adjustment required), when hour (maximum adjustment range), the intermediate values are determined by linear interpolation, that is, (when This factor is used to make a weighted correction to the current level reasoning interval. The corrected interval value for: ; in, is the hierarchical reasoning interval before correction (unit: seconds), is the dynamic adjustment factor (dimensionless).

[0053] Taking knowledge base X as an example, its initial level reasoning interval Seconds, preset duration Seconds, the actual effective time of a certain time in the historical records seconds, the deviation ratio , dynamic adjustment factor , the corrected interval Second.

[0054] After completing the hierarchical reasoning interval correction of all aggregated knowledge bases, the system summarizes the weighted hierarchical reasoning intervals , rearrange the reasoning priority sequence The sorting rule is still from small to large interval values. For example, the original initial sequence is , the interval of the corrected knowledge base X becomes seconds, and the initial interval of another knowledge base Y is Seconds, corrected , , seconds, the new sequence is , the corresponding reasoning order is adjusted to knowledge base A→knowledge base Y→knowledge base X.

[0055] Adjusted inference priority sequence This information is synchronized with the path construction of subsequent inference nodes, ensuring that the entire inference process is executed based on real-time optimized rules. For example, when building the next inference node, the system prioritizes inference calls from earlier knowledge bases in the sequence to improve inference efficiency and accuracy.

[0056] For the baseline knowledge base of decentralized features, the inference rule is to initiate inference in an appended manner after the inference of the aggregated knowledge base is completed. This appended inference uses a parallel processing mode, which means that multiple decentralized knowledge bases are simultaneously invoked to cross-validate or supplement the current inference results. For example, after the aggregated inference of architectural design is completed, decentralized knowledge bases such as structural safety specifications and energy-saving design standards are activated to conduct parallel assessments of the safety and environmental performance of the design solution, generating multi-dimensional inference results.

[0057] The system also incorporates a conflict detection mechanism when executing inference rules. When conflicting inference results emerge from different knowledge bases (e.g., a converged knowledge base recommends a certain material type, while a decentralized knowledge base suggests another), the system automatically triggers a weighted evidence comparison. It assigns weights to each inference result based on factors such as the authority and timeliness of the knowledge base, prioritizing results with higher weights. If the weights are identical, a conflict report is generated, prompting the design consultant to intervene in the decision-making process.

[0058] Furthermore, the dynamic adjustment process of inference rules is fully documented. The system records information such as the time of each adjustment, the object of adjustment, the interval value before and after the adjustment, the deviation ratio, and the adjustment factor, forming an audit log. This audit log can be used to trace the evolution of inference rules, facilitating subsequent optimization and debugging of inference models.

[0059] Example 4: The system uses a mechanical component design inference node as an example to verify the validity of each inference node and calculate deviations under abnormal conditions. This inference node's task is to perform inference based on the user's requirement to "design the parameters of a transmission system gear." The node draws on a material knowledge base, a design specification knowledge base, and a historical case knowledge base to produce the output: "Gear material selection: 45 steel, module m = 3, number of teeth z = 20, tooth width b = 60 mm."

[0060] First, the inference results are transformed into vectors. The system extracts semantic entities and relational attributes from the inference results: semantic entities include "45 steel," "module m=3," "number of teeth z=20," and "tooth width b=60mm." Relational attributes include "material-performance matching" (matching of 45 steel with gear load requirements), "parameter-design specification association" (compliance of the module and number of teeth with mechanical design manual standards), and "parameter-case experience association" (comparison of tooth width values with historical similar cases). Based on these entities and attributes, a multidimensional feature vector is constructed, with the vector dimensions corresponding to material type, module, number of teeth, tooth width, and various associations, such as [material, module, number of teeth, tooth width, material-performance matching, parameter specification compliance, parameter-case association].

[0061] Perform dimensionality reduction on multidimensional feature vectors. Because high-dimensional vectors may contain redundant information, the system uses the principal component analysis (PCA) algorithm to map the data in the original high-dimensional space to a low-dimensional vector space through linear transformation, while preserving the key feature information. Assuming the original vector is 7-dimensional, after dimensionality reduction, it is mapped to a 2-dimensional space, resulting in a standardized mapping result (0.6, 0.7), which reflects the comprehensive performance of the inference results in key feature dimensions.

[0062] The standardized mapping results are weighted according to a preset semantic weight matrix. This semantic weight matrix is pre-set based on the importance of knowledge within the design consulting field. For example, the weight for material performance matching is 40%, for parameter specification compliance is 30%, and for parameter case relevance is 30%. The reduced-dimensional mapping results (0.6, 0.7) are mapped to the principal components of the weight matrix. Assuming the first principal component corresponds to the combined characteristics of material performance and parameter specifications (weighted 70%), and the second principal component corresponds to the parameter case characteristics (weighted 30%), the weighted calculation process is: verification vector value = 0.6 × 0.7 + 0.7 × 0.3 = 0.63. This value serves as the final vector value of the input verification function.

[0063] The preset verification function is called to evaluate the verification vector. This verification function uses a support vector machine (SVM) classification model, trained with historical inference data, to distinguish between normal and abnormal inference results. After the verification vector (0.63) is input into the function, the output verification parameter is a probability value of 0.6, which represents the probability that the inference result is normal. The system's preset verification threshold is 0.7. Since 0.6 < 0.7, the inference node is considered to be in a normal inference state, allowing subsequent inferences to continue in a hierarchical order, such as triggering the gear strength verification knowledge base to perform mechanical analysis on the current parameters.

[0064] If an inference node outputs an anomaly, for example, another mechanical design node, whose inference result is "Gear material selection: 20 steel, module m = 2, number of teeth z = 15, tooth width b = 30 mm," is extracted, along with the semantic entities "20 steel," "module m = 2," "number of teeth z = 15," and "tooth width b = 30 mm," along with the relationship attributes "Material-Performance Matching" (20 steel has low strength and may not meet load-bearing requirements) and "Parameter-Design Specification Association" (module 2 is less than the industry-recommended minimum of 2.5). After constructing a multidimensional feature vector, dimensionality reduction is performed to obtain a standardized mapping result (0.4, 0.9). Combined with the semantic weight matrix (material performance weighting 40%, specification compliance weighting 30%), the verification vector value is calculated as 0.4 × 0.7 + 0.9 × 0.3 = 0.55. After inputting the SVM verification function, the output verification parameter is 0.8, which is greater than the preset threshold of 0.7, indicating an abnormal inference state.

[0065] Entering the abnormal state processing flow, the system first calculates the reasoning deviation of the baseline knowledge base at the abnormal reasoning node. The knowledge bases located for incomplete associated reasoning are the materials knowledge base and the design specification knowledge base. The actual reasoning node is the current abnormal node (step 3). According to the reasoning rules, reasoning for the materials knowledge base should be completed in step 1, and for the design specification knowledge base in step 2. Therefore, the logical intervals between the actual reasoning node and the abnormal node are: the materials knowledge base interval = 3-1 = 2 steps, and the design specification knowledge base interval = 3-2 = 1 step. The maximum value of 2 is taken as the real-time deviation.

[0066] Perform historical backtracking on abnormal reasoning nodes and extract the historical deviations of their associated historical nodes. Assume that historical records show that the deviations for similar material selection problems in the previous two reasonings were 1 step and 1.5 steps, respectively. The system's preset fault tolerance threshold is 2 steps. Since 1 ≤ 2 and 1.5 ≤ 2, if we continue backtracking to earlier historical nodes and find that the deviation for the previous one was 2.5 steps (> 2), we stop backtracking and finally extract the historical deviations of 1 step and 1.5 steps for the most recent two times.

[0067] Call the deviation calculation function (such as taking the arithmetic mean), input the real-time deviation of 2 and the historical deviations of 1 and 1.5 into the function, and calculate the inference deviation = (2 + 1 + 1.5) / 3 ≈ 1.5. This deviation reflects the degree of deviation of the abnormal node from the inference path, providing a quantitative basis for subsequent path correction.

[0068] Throughout the verification and deviation calculation process, the system stores structured input and output data for each step, including the original inference result data, multidimensional feature vectors, the resulting mapping after dimensionality reduction, the verification vector value, verification parameters, and the deviation calculation process. This data can be used to subsequently optimize the inference model training. For example, by analyzing the feature vector distribution of abnormal nodes and adjusting the semantic weight matrix or verification function parameters, the system can improve its accuracy in identifying abnormal conditions.

[0069] In addition, the system supports dynamic updates to the validation function. After accumulating a certain number of abnormal reasoning cases, the model retraining process is automatically triggered. The SVM classifier is retrained using the newly added data, updating the classification boundaries to ensure that the validation function can adapt to emerging reasoning patterns or changes in data distribution.

[0070] Example 5: During the path correction process for abnormal reasoning nodes, the system uses a specific reasoning scenario for building structure design as an example. For example, in a frame structure column cross-section dimensioning reasoning node, the system initially accesses the building load code knowledge base, the concrete structure design code knowledge base, and the historical case knowledge base, outputting "column cross-section dimensions 400mm x 400mm, concrete strength grade C30." However, validity verification indicates an abnormal reasoning state (verification parameter 0.85 ≥ a preset threshold of 0.7).

[0071] The system first counts the number of historical backtrackings for the abnormal reasoning node. Assuming this is the first time an anomaly has been triggered at this node, the historical backtracking count is 0, and the adjustment reference value is recorded as 0. The system presets an adjustment threshold of 3. Since 0 < 3, the system enters the continuous collection of inference deviations. At this point, the system recalculates the real-time deviation: the knowledge base for the incomplete associated reasoning is the seismic design code knowledge base (because the seismic fortification intensity in the area is 7 degrees, verification of the column axial compression ratio is required), and the actual inference node is the current abnormal node (step 4). Seismic design code reasoning should be performed after load and concrete code reasoning and before case-based reasoning (i.e., step 3). Therefore, the logical interval is 4 - 3 = 1 step, which is used as the real-time deviation. Simultaneously, a backtracking of the historical nodes reveals that this type of anomaly has never occurred in historical reasoning, resulting in a historical deviation of 0. Therefore, the inference deviation equals the real-time deviation of 1. Since the adjustment reference value does not reach the threshold, the system does not immediately trigger path correction. Instead, it continues to monitor the subsequent reasoning process and accumulates deviation data.

[0072] When the abnormal inference node triggers an anomaly for the second time, the historical backtracking count is updated to 1. The adjustment reference value is still less than the adjustment threshold of 3, and the system continues to collect deviations. Assuming the second real-time deviation is 1.5 steps (because the newly added wind load code knowledge base is not used), the historical deviation is accumulated to 1 step of the previous one, and the inference deviation is calculated as (1 + 1.5) / 2 = 1.25. At this point, the system still does not perform corrections, but instead stores the two deviation data in the historical record for analysis of abnormal patterns.

[0073] When the abnormal reasoning node triggered an anomaly for the third time, the number of historical backtrackings reached three (equal to the adjustment threshold), triggering the path correction process. The real-time deviation was now two steps (due to a misjudgment of the structural type, resulting in a mismatch in the order of multiple knowledge base invocations), while the historical deviations were 1 and 1.5 steps from the previous two inferences, resulting in a reasoning deviation of (2 + 1 + 1.5) / 3 ≈ 1.5. The system then logically offset the abnormal reasoning node based on the inference deviation. First, the system analyzed the source of the deviation and found that the missing parameters were due to the failure to prioritize the seismic design code knowledge base. Therefore, a correction reasoning node was generated, adjusting the reasoning order to: building load code knowledge base → seismic design code knowledge base → concrete structure design code knowledge base → historical case knowledge base. Within the correction reasoning node, the associated reasoning was re-executed: based on a seismic fortification intensity of 7, the seismic design code knowledge base was used to calculate the column axial compression ratio limit. Combined with the axial force design value derived from the load code, the column cross-section dimensions were recalculated to 500 mm × 500 mm, with the concrete strength grade maintained at C30. The validity was verified again, and the semantic entities "500mm×500mm" and "C30" and the relationship attributes "seismic verification compliance" and "axial compression ratio rationality" were extracted. After constructing a multi-dimensional feature vector and performing dimensionality reduction processing, the verification vector value was calculated to be 0.68. The input verification function output parameter 0.65<0.7 was determined to be a normal reasoning state, and subsequent reasoning steps such as floor stiffness ratio calculation were continued.

[0074] If the system detects a significant change in the deviation pattern of an abnormal inference node (e.g., a shift from a parameter-missing anomaly to a rule-conflicting anomaly) before the adjustment reference value reaches the threshold, it automatically resets the historical backtracking count to avoid invalid corrections due to differences in anomaly types. For example, if a mechanical design anomaly node was triggered twice by material parameter conflicts (with historical deviations of 1.2 and 1.4, respectively), and the third time, due to an unsynchronized change in user requirements, resulted in a deviation from the inference target (a real-time deviation of 2.0), the system would identify this as a new anomaly type, reset the historical backtracking count, and reaccumulate the adjustment reference value.

[0075] During the path correction process, the system's logical offset operation is based on a preset backup rule base. This backup rule base contains alternative reasoning paths or supplementary rules corresponding to each benchmark knowledge base. For example, when the original design code rules conflict with case experience, the system can switch to industry-recommended empirical formula rules or expert consensus rules. For example, in bridge design, if the original reasoning rules trigger an exception due to bridge span parameters exceeding the applicable range of the code, the logical offset can invoke the rules in the special design guide for long-span bridges and generate a corrected reasoning node "Use finite element analysis instead of simplified calculations."

[0076] The system keeps a detailed record of each path correction operation, including correction time, abnormal node identification, reasoning deviation, adjustment reference value, and comparison of the reasoning paths before and after the correction, creating a correction log. This log allows design consultants to trace the reasoning process and analyze the causes of abnormalities. It also provides data support for optimizing the structure of the benchmark knowledge base or the reasoning rules. For example, by counting the high frequency of a certain type of abnormality, it is possible to determine whether there are missing knowledge items or conflicting rules in the benchmark knowledge base, thereby triggering the knowledge base update process.

[0077] Furthermore, path correction supports manual intervention. When a system-generated correction inference node presents multiple possible solutions (e.g., different backup rules lead to different correction paths), a prompt box will pop up for the design consultant to select, ensuring that the correction meets actual project requirements. For example, when correcting an abnormality in a building energy-saving design, the system might offer two correction options: "adjust the thickness of the exterior wall insulation material" or "change the airtightness level of the window type." The designer can then make a decision based on project cost and deadline requirements.

[0078] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0079] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A knowledge reasoning method for design consulting, characterized by: include: Acquire diversified knowledge data for design consulting needs, perform knowledge preprocessing, and generate a benchmark knowledge base; Extracting knowledge association features of each of the benchmark knowledge bases, and determining inference rules for different knowledge bases based on the knowledge association features; Constructing reasoning paths for each benchmark knowledge base according to the reasoning rules to form multiple groups of reasoning chains and reasoning nodes for each of the reasoning chains; Performing validity verification on each of the inference nodes to obtain an inference state of each node, wherein the inference state includes a normal inference state and an abnormal inference state; In the normal reasoning state, the reference knowledge bases under each reasoning node perform associative reasoning in a hierarchical order; In the abnormal reasoning state, identifying the abnormal reasoning node and calculating the reasoning deviation of the benchmark knowledge base under the abnormal reasoning node; Based on the inference deviation, the path of the abnormal inference node is corrected until all the benchmark knowledge bases complete the inference in hierarchical order and terminate.

2. The knowledge reasoning method for design consultation according to claim 1, characterized in that: The acquisition of diversified knowledge data of design consulting needs and knowledge preprocessing to generate a benchmark knowledge base includes: Collect design requirement documents, historical case data, and industry specification data, integrate them into independent knowledge units, and assign a unique identifier to each of the knowledge units; Performing redundancy filtering on the data within each of the independent knowledge units to remove duplicate or conflicting knowledge entries; The filtered independent knowledge units are semantically standardized to unify the data structures of the knowledge units, and the structurally unified independent knowledge units are output as a benchmark knowledge base.

3. The knowledge reasoning method for design consultation according to claim 1, characterized in that: The extracting of knowledge association features of each of the benchmark knowledge bases includes: Obtaining the knowledge nodes under each of the benchmark knowledge bases and sorting them according to the strength of association; Extracting semantic relevance between adjacent knowledge nodes and marking it as classification condition parameter; Obtaining a preset correlation threshold, dividing each classification condition parameter into intervals according to the correlation threshold, and generating multiple correlation intervals; Counting the number of classification condition parameters within each of the association intervals and recording them as feature distribution parameters, and determining the knowledge association features of each benchmark knowledge base based on the feature distribution parameters; The knowledge association features include aggregation features and dispersion features, and the reasoning priority of the dispersion features is lower than the reasoning priority of the aggregation features.

4. The knowledge reasoning method for design consultation according to claim 3 is characterized by: The determining of the knowledge association features of each benchmark knowledge base based on the feature distribution parameters includes: Obtaining feature distribution parameters under each of the benchmark knowledge bases; Sort the feature distribution parameters under the same benchmark knowledge base by value from high to low, and calculate the weight of the feature distribution parameter with the highest value; Obtaining a weight threshold, and comparing the weight threshold with the weight of the characteristic distribution parameter with the highest value; If the proportion weight is greater than the weight threshold, it is determined that the reference knowledge base corresponding to the feature distribution parameter has an aggregation feature; If the proportion weight is less than or equal to the weight threshold, it is determined that the reference knowledge base corresponding to the feature distribution parameter has a dispersion feature.

5. The knowledge reasoning method for design consultation according to claim 3 is characterized by: The determining of inference rules of different knowledge bases based on the knowledge association features includes: Acquiring knowledge association features of each of the benchmark knowledge bases; The maximum value of the correlation interval corresponding to the aggregation feature is used as the hierarchical reasoning interval; Summarize the hierarchical reasoning intervals of the benchmark knowledge base under all the aggregation features, and arrange them from small to large according to the interval values to generate a reasoning priority sequence; Performing hierarchical reasoning on each of the benchmark knowledge bases according to the reasoning priority sequence, and after the reasoning of the benchmark knowledge base under the aggregation feature is completed, additional reasoning is performed on the benchmark knowledge base under the dispersion feature; Generating the reasoning priority sequence includes: Obtaining historical adjustment records of hierarchical reasoning intervals of a benchmark knowledge base under the aggregation feature; Extract the actual effective duration of the hierarchical reasoning interval in the historical reasoning path and calculate the deviation ratio between it and the preset duration; Generating a dynamic adjustment factor based on the deviation ratio, and performing a weighted correction on the current hierarchical reasoning interval according to the adjustment factor; Summarize the weighted hierarchical reasoning intervals, rearrange the reasoning priority sequence, and synchronize the adjusted sequence to the path construction of subsequent reasoning nodes.

6. The knowledge reasoning method for design consultation according to claim 1, characterized in that: The performing validity verification on each of the inference nodes includes: Obtaining the inference results of the benchmark knowledge base under each of the inference nodes, and performing vector mapping conversion to generate multiple verification vectors; Calling a preset verification function, inputting the verification vector into the verification function, and recording the output value as a verification parameter; Obtaining a verification threshold, and comparing the verification parameter with the verification threshold; If the verification parameter is less than the verification threshold, it is determined that the inference node is in a normal inference state; If the verification parameter is greater than or equal to the verification threshold, it is determined that the inference node is in an abnormal inference state.

7. The knowledge reasoning method for design consultation according to claim 1 is characterized by: The calculating of the reasoning deviation of the benchmark knowledge base at the abnormal reasoning node includes: Obtaining an actual reasoning node of a benchmark knowledge base for which associated reasoning has not been completed under the abnormal reasoning node; Calculating the logical interval between the actual reasoning node and the abnormal reasoning node, and calibrating it as a real-time deviation; Perform historical backtracking on the abnormal reasoning node and extract the historical deviation of its associated historical nodes; Obtaining a preset fault tolerance threshold, and terminating the backtracking operation when the historical deviation is less than or equal to the fault tolerance threshold; The deviation measurement function is called, the real-time deviation amount and the historical deviation amount are input into the measurement function, and the output result is recorded as the inferred deviation amount.

8. The knowledge reasoning method for design consultation according to claim 7, characterized in that: The performing path correction on the abnormal reasoning node based on the reasoning deviation includes: Count the historical backtracking times of the abnormal reasoning node and record it as an adjustment reference value; Obtaining an adjustment threshold, and comparing the adjustment reference value with the adjustment threshold; When the adjustment reference value is greater than or equal to the adjustment threshold, performing a logical shift on the abnormal reasoning node according to the reasoning deviation amount, generating a corrected reasoning node, and performing associated reasoning under the corrected reasoning node; When the adjustment reference value is less than the adjustment threshold, the inference deviation is continuously collected until the adjustment reference value reaches or exceeds the adjustment threshold, and then the path correction is triggered.

9. The knowledge reasoning method for design consultation according to claim 6, characterized in that: The vector mapping conversion includes: Extract semantic entities and relationship attributes from the reasoning results and construct multidimensional feature vectors; Performing dimensionality reduction processing on the multidimensional feature vector to generate a standardized mapping result in a low-dimensional vector space; The standardized mapping result is weightedly calculated according to a preset semantic weight matrix to generate the verification vector and input it into the verification function.

10. A knowledge reasoning system for design consulting, characterized by: The knowledge reasoning method for design consultation according to any one of claims 1 to 9 comprises: A data acquisition module, which is used to acquire diversified knowledge data of design consulting needs, perform knowledge preprocessing, and generate a benchmark knowledge base; A rule output module, the rule output module is used to extract the knowledge association features of each of the benchmark knowledge bases and determine the inference rules of different knowledge bases based on the knowledge association features; An inference generation module, configured to construct an inference path for each benchmark knowledge base according to the inference rules, forming multiple groups of inference chains and inference nodes for each of the inference chains; A state evaluation module, the state evaluation module is used to verify the validity of each of the inference nodes to obtain the inference state of each node; In the normal reasoning state, the reference knowledge bases under each reasoning node perform associative reasoning in a hierarchical order; In the abnormal reasoning state, identifying the abnormal reasoning node and calculating the reasoning deviation of the benchmark knowledge base under the abnormal reasoning node; The inference optimization module is used to correct the path of the abnormal inference node based on the inference deviation until all the benchmark knowledge bases complete the inference in hierarchical order and terminate.

Images