Abnormal working condition data analysis and early warning method and system for drying machine

Abstract

The embodiment of the invention provides an abnormal working condition data analysis and early warning method and system for a drying machine, and the method comprises the steps: firstly constructing a diagnosis knowledge graph of the abnormal working condition of the drying machine, which comprises entity types of drying machine parts, operation parameters, abnormal working condition types, environmental factors and the like; then, working condition parameter data collected by a drying machine in real time is subjected to correlation mapping with the working condition parameter data, knowledge graph instantiation data are generated, multi-hop correlation reasoning is carried out according to a reasoning rule set in the diagnosis knowledge graph, and the diagnosis knowledge graph instantiation data is obtained; and constructing an abnormal association path set containing entity nodes, relation edges and attribute values, screening out a target abnormal association path of which the comprehensive confidence exceeds a preset threshold value from the abnormal association path set, generating dryer abnormal working condition early warning information according to information contained in the target abnormal association path, and sending the dryer abnormal working condition early warning information to a dryer monitoring terminal. Therefore, accurate analysis and timely early warning of the abnormal working condition of the drying machine are realized.

Description

Technical Field

[0001] The present application relates to the technical field of dryer operation monitoring based on knowledge reasoning, and in particular to a method and system for analyzing and warning abnormal operating condition data of a dryer. Background Art

[0002] During dryer operation, various abnormal operating conditions often occur due to factors such as complex equipment, volatile operating environments, and varying material properties. If these abnormal conditions are not promptly detected and addressed, they will not only affect the dryer's normal operation and reduce production efficiency, but may also cause equipment damage, increase repair costs, and even lead to safety accidents. Currently, monitoring and analyzing abnormal dryer conditions primarily relies on traditional fault diagnosis methods. Some methods set fixed parameter thresholds, triggering alarms when real-time collected parameters exceed the threshold range. However, these methods are overly simplistic and fail to account for the complex correlations between parameters and the dynamic changes under different operating conditions, making them prone to false positives and missed negatives. Other methods utilize model-based analysis, but establishing an accurate dryer model is difficult, and the model lacks adaptability to factors such as equipment aging and environmental changes, resulting in low diagnostic accuracy. Furthermore, most existing technologies lack in-depth analysis of the abnormal condition's propagation path, making it difficult to quickly locate the root cause of the fault and predict its development trend, thus failing to provide comprehensive and accurate early warning information to operators. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method and system for analyzing and warning abnormal operating condition data of a dryer.

[0004] According to a first aspect of the present application, a method for analyzing and warning abnormal operating condition data of a dryer is provided, the method comprising: Construct a diagnostic knowledge graph for abnormal operating conditions of a dryer. The diagnostic knowledge graph includes an entity type set, a relationship type set, an attribute set, and an inference rule set. The entity type set includes dryer component entities, operating parameter entities, abnormal operating condition type entities, and environmental factor entities. The relationship type set includes the association relationship between components and parameters, the representation relationship between parameters and operating conditions, and the impact relationship between operating conditions and the environment. The attribute set includes feature description information corresponding to each entity type. The inference rule set includes abnormality propagation logic rules based on the relationship between entities. Performing an association mapping process on the operating parameter data collected in real time by the dryer and the diagnostic knowledge graph, establishing an association relationship between the real-time operating parameter data and the attribute values ​​of the corresponding entities, and generating knowledge graph instantiation data containing the real-time attribute values; Based on the inference rule set, multi-hop association reasoning is performed on the instantiated data of the knowledge graph, and the dryer component entity, environmental factor entity and abnormal operating condition type entity associated with the real-time operating condition parameter data are retrieved, and an abnormal association path set including entity nodes, relationship edges and attribute values ​​is constructed; Calculate the confidence of each abnormal associated path in the abnormal associated path set, determine the comprehensive confidence of the path according to the weight value of each relationship edge in the path and the abnormal deviation of the entity attribute, and screen out the target abnormal associated paths whose comprehensive confidence exceeds a preset threshold; Dryer abnormal operating condition warning information is generated according to the abnormal operating condition type entity, dryer component entity and abnormal propagation relationship included in the target abnormality association path, and the abnormal operating condition warning information is sent to the dryer monitoring terminal.

[0005] According to the second aspect of the present application, a data analysis and early warning system for abnormal operating conditions of a dryer is provided. The data analysis and early warning system for abnormal operating conditions of a dryer includes a processor and a readable storage medium. The readable storage medium stores a program, and when the program is executed by the processor, it implements the aforementioned data analysis and early warning method for abnormal operating conditions of the dryer.

[0006] Based on any of the above aspects, the present application constructs a diagnostic knowledge graph containing multiple entity types, relationship types, attribute sets and inference rule sets, comprehensively and systematically covers all kinds of key information involved in the operation of the dryer, associates and maps the real-time collected operating parameter data with the knowledge graph, generates knowledge graph instantiation data containing real-time attribute values, so that the knowledge graph can reflect the actual operating status of the dryer in real time, performs multi-hop association reasoning based on the inference rule set, and constructs an abnormal association path set, which can deeply explore the propagation relationship between abnormal working conditions between different entities, accurately find out the components and environmental factors related to the abnormality, and effectively improve the accuracy and reliability of abnormal judgment by calculating the confidence of the abnormal association path and screening out the target abnormal association path, avoiding false alarms and missed alarms. Finally, abnormal working condition warning information is generated and sent according to the target abnormal association path, providing operators with comprehensive, accurate and timely warnings, which helps to quickly locate the root cause of the fault, predict the abnormal development trend, and take timely measures to ensure the safe and stable operation of the dryer, improve production efficiency and reduce maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 A flow chart of a method for analyzing and warning abnormal operating condition data of a dryer provided in an embodiment of the present application is shown; Figure 2 A schematic diagram of the component structure of the abnormal operating condition data analysis and early warning system for a dryer provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0008] Figure 1 The following is a flow chart illustrating a method for analyzing and warning abnormal operating conditions of a dryer, provided in an embodiment of the present application. It should be understood that in other embodiments, the order of some steps in this method may be interchanged, or some steps may be omitted or deleted, depending on actual needs. The detailed steps of this method are described below.

[0009] Step S110: Construct a diagnostic knowledge graph of abnormal operating conditions of the dryer, wherein the diagnostic knowledge graph includes an entity type set, a relationship type set, an attribute set, and an inference rule set. The entity type set includes dryer component entities, operating parameter entities, abnormal operating condition type entities, and environmental factor entities. The relationship type set includes the association relationship between components and parameters, the representation relationship between parameters and operating conditions, and the influence relationship between operating conditions and the environment. The attribute set includes feature description information corresponding to each entity type. The inference rule set includes abnormal propagation logic rules based on the relationship between entities.

[0010] In this embodiment, in order to accurately analyze and warn of abnormal operating conditions of the dryer, it is necessary to construct its diagnostic knowledge graph. The diagnostic knowledge graph is a structured knowledge representation method that integrates various information during the operation of the dryer to facilitate subsequent analysis and reasoning.

[0011] Step S111: collecting dryer design documents, operation and maintenance manuals, and historical fault case data, and extracting dryer component names, operation parameter names, abnormal operating condition type names, and environmental factor names as entity candidate sets.

[0012] Data collection is fundamental to building a diagnostic knowledge graph. Dryer design documentation details the dryer's internal structure, operating principles, and technical parameters. The operation and maintenance manual includes the dryer's daily operating specifications, maintenance requirements, and troubleshooting methods for common faults. This information is extremely helpful for determining the dryer's normal operating status and identifying abnormalities. Historical failure case data is a detailed record of past dryer failures, including the time, symptoms, and handling process. By analyzing this data, common abnormal operating conditions and causes of dryer failures can be identified.

[0013] Entity candidate sets are extracted from this data. Dryer design documents contain the names of various dryer components, such as heating elements, ventilation ducts, and material conveying devices. The heating element is the core component of the dryer that generates heat, the ventilation ducts facilitate air circulation, and the material conveying device transports the material to be dried into the dryer. Operation and maintenance manuals mention operating parameters such as temperature, humidity, and pressure. Temperature is a key parameter in the drying process; maintaining an appropriate temperature range ensures effective drying. Humidity reflects the moisture content of the air inside the dryer and has a significant impact on the drying rate. Pressure affects the airflow conditions within the dryer. Historical fault case data includes the names of abnormal operating conditions, such as overheating, blockage, and leakage, as well as environmental factors such as ambient temperature and humidity. Ambient temperature and humidity affect the dryer's operating efficiency and stability, and dryers may exhibit different abnormal operating conditions under different environmental conditions.

[0014] Step S112: labeling the entity candidate set with entity types, dividing the entity candidate set into dryer component entities, operating parameter entities, abnormal operating condition type entities, and environmental factor entities to form an entity type set.

[0015] Labeling and categorizing entity candidates helps organize and manage knowledge more clearly. For dryer component entities, we identify them based on their physical function and structural characteristics within the dryer. For example, the heating element, ventilation duct, and material transfer device mentioned above are specific components of the dryer, with clear physical form and function. Therefore, they can be labeled as dryer component entities.

[0016] The criteria for identifying an operating parameter entity is whether it describes the dryer's operating status. Parameters like temperature, humidity, and pressure directly reflect the dryer's various operating conditions. These parameters change continuously as the dryer operates, so these parameters are labeled as operating parameter entities.

[0017] Abnormal operating condition entities are identified based on whether they represent an abnormal condition in the dryer. Overheating means the dryer's temperature exceeds the normal range, potentially damaging dryer components. Blockages can affect material transfer and air circulation, causing the dryer to malfunction. Leaks can waste energy and pollute the environment. These abnormal conditions can all be labeled as abnormal operating condition entities.

[0018] Environmental factor entities are factors related to the dryer's operating environment. Ambient temperature and humidity can affect dryer operation. For example, in high temperatures, the dryer's heat dissipation becomes more difficult, making it prone to overheating. In humid environments, internal dryer components may rust, affecting normal operation and potentially causing abnormal operating conditions such as component damage and blockage. Therefore, factors such as ambient temperature and humidity are labeled as environmental factor entities. Through these labeling and classification, a set of entity types is formed.

[0019] Step S113: Based on the historical fault case data and the operation and maintenance manual, the association relationship between different entity types is analyzed, and the association relationship between components and parameters, the representation relationship between parameters and working conditions, and the impact relationship between working conditions and the environment are defined to form a relationship type set.

[0020] Through in-depth research on historical failure case data and operation and maintenance manuals, it can be found that there is a close correlation between different entity types.

[0021] Regarding the relationship between components and parameters, for example, the operating status of the heating element directly affects the temperature parameters within the dryer. If the heating element malfunctions, such as a power loss or damage, the temperature within the dryer will not reach the set value, resulting in poor drying results. Blockage in ventilation ducts can affect air circulation, which in turn affects the pressure and humidity parameters within the dryer. When the ventilation ducts are blocked, air cannot circulate properly, resulting in increased pressure and humidity due to the inability to drain moisture in a timely manner.

[0022] The relationship between parameters and operating conditions is the basis for identifying abnormal operating conditions by observing changes in operating parameters. A sustained rise in temperature parameters beyond the normal range could indicate overheating of the dryer. Sudden, abnormal fluctuations in pressure parameters could indicate a blockage or leak within the dryer. By analyzing numerous historical failure cases, we have established relationships between abnormal parameter changes and various abnormal operating conditions.

[0023] The relationship between operating conditions and the environment reflects the impact of the external environment on abnormal dryer operation. High temperatures increase the dryer's heat dissipation burden, making overheating more likely. High ambient temperatures reduce the temperature difference between the dryer and the outside world, slowing heat dissipation and causing heat accumulation inside the dryer, leading to elevated temperatures. In humid environments, dryer components are prone to rusting, impacting normal operation and potentially causing abnormal operating conditions such as component damage and blockage. Through analysis and summary of these relationships, a set of relationship types was formed.

[0024] Step S114: define a set of attributes for each entity type, where the attributes of the dryer component entity include component model, installation location, and normal operating range; the attributes of the operating parameter entity include parameter unit, sampling frequency, and normal value range; the attributes of the abnormal operating condition type entity include fault feature description, occurrence probability, and impact degree; and the attributes of the environmental factor entity include factor category, measurement range, and impact threshold.

[0025] The purpose of defining a set of attributes for an entity type is to describe and distinguish each entity in more detail. For dryer component entities, the component model is used to uniquely identify the specifications and performance of the component. Different models of components may differ in power, size, material, etc., which is very important for the selection and replacement of components. The installation position clarifies the specific location of the component in the dryer, making it easier for maintenance personnel to quickly locate and repair it. The normal operating range specifies the range of various parameters when the component is operating normally, such as the normal operating temperature range of the heating element, the normal pressure range of the ventilation duct, etc. If the range is exceeded, the component may malfunction.

[0026] In the properties of the operating parameter entity, the parameter unit is used to unify the parameter's measurement standard. For example, the temperature unit can be a standard measurement unit, the humidity unit can be a proportional measurement unit, and the pressure unit can be a pressure measurement unit. The sampling frequency determines how often parameter data is collected. A higher sampling frequency can reflect parameter changes more promptly, but it also increases the data processing burden; a lower sampling frequency may miss some important parameter changes. The normal value range is an important basis for determining whether a parameter is abnormal. When a parameter exceeds this normal value range, the operating status of the dryer needs to be paid attention to.

[0027] In the attributes of the abnormal operating condition type entity, the fault feature description details the specific manifestations of the abnormal operating condition. For example, the fault characteristics of overheating may be excessively high surface temperature of the dryer or an odor. The probability of occurrence is derived from the statistics of historical failure case data, which can help predict the possibility of a certain abnormal operating condition in the dryer. The impact level assesses the degree of harm caused by the abnormal operating condition to the operation and production of the dryer. For example, overheating may cause damage to dryer components and deteriorate material quality.

[0028] In the attributes of the environmental factor entity, the factor category specifies the specific type of environmental factor, such as temperature or humidity. The measurement range specifies the range of possible values ​​for the environmental factor in actual measurements, and the impact threshold determines whether the environmental factor has an adverse impact on dryer operation. When the environmental factor exceeds the impact threshold, the probability of the dryer operating abnormally increases.

[0029] Step S115: Based on the entity type set, relationship type set and attribute set, a graph database is used to construct the basic structure of the diagnostic knowledge graph, wherein entities are nodes of the graph, relationships are edges connecting nodes, and attributes are stored as attribute values ​​of nodes.

[0030] A graph database is a database specifically designed for processing graph-structured data and is well-suited for building knowledge graphs. When constructing the infrastructure for a diagnostic knowledge graph, each entity is considered a node in the graph. For example, the heating element, ventilation duct, and material transfer device in the dryer component entity each become a node; the temperature, humidity, and pressure in the operating parameter entity are also nodes; the overheating, blockage, and leakage in the abnormal operating condition type entity are also nodes; and the ambient temperature and humidity in the environmental factor entity are also represented as nodes.

[0031] The relationship between different entities is represented by edges connecting nodes. The association relationship between components and parameters can be represented by edges connecting the dryer component entity node and the operating parameter entity node. For example, the heating element node and the temperature node are connected by an edge, indicating that the working state of the heating element will affect the temperature parameter. The representation relationship between parameters and working conditions is represented by edges connecting the operating parameter entity node and the abnormal working condition type entity node. For example, the edge between the temperature node and the overheating node indicates that an abnormal temperature increase may lead to an abnormal working condition of overheating. The impact relationship between working conditions and the environment is represented by edges connecting the abnormal working condition type entity node and the environmental factor entity node. For example, the edge between the overheating node and the ambient temperature node indicates that an excessively high ambient temperature increases the probability of an abnormal overheating working condition.

[0032] The attributes of each node are stored as their attribute values ​​in the graph database. For example, the attribute values ​​of a heating element node might include the component model, installation location, and normal operating range; the attribute values ​​of a temperature node might include the parameter unit, sampling frequency, and normal value range. This approach builds the foundational structure of a diagnostic knowledge graph.

[0033] Step S116: Extract the abnormal propagation logic from the historical fault case data and convert it into a set of inference rules of the diagnostic knowledge graph. The inference rule set includes a rule antecedent and a rule consequent. The rule antecedent is the condition satisfied by the entity attribute or the relationship between entities, and the rule consequent is the new relationship obtained by reasoning or the abnormal working condition type judgment result.

[0034] In this embodiment, the anomaly propagation logic is extracted from historical fault case data in order to enable the knowledge graph to perform intelligent reasoning.

[0035] Step S1161: Structural processing is performed on historical fault case data. Each fault case includes the time of fault occurrence, a list of involved components, a list of abnormal parameters, environmental factor records, and fault type diagnosis results.

[0036] Historical failure case data is usually in a rather messy format, and structured processing is to organize it into a unified and standardized format. Recording the time when the failure occurred helps to analyze the patterns and trends of failures and understand the differences in the frequency of failures in different time periods. The list of involved parts clearly states which dryer components were affected when the failure occurred, which is critical for locating the source of the failure. The abnormal parameter list records the abnormal changes in operating parameters. Abnormalities in these parameters are often important signals of failure. Environmental factor records include information such as the ambient temperature and humidity at the time. Because environmental factors may affect the operation of the dryer, the possibility and type of dryer failure may be different under different environmental conditions. The fault type diagnosis result is a clear judgment of the fault.

[0037] Step S1162: Use natural language processing technology to perform entity recognition and relationship extraction on the text description of the fault case, identify the dryer component entity, operating parameter entity, abnormal operating condition type entity and environmental factor entity contained in the case, and extract the association relationship between the entities and the abnormality of the attribute value.

[0038] Natural language processing technology can extract useful information from the text descriptions of fault cases. Entity recognition can accurately locate the various entities involved in the case, for example, accurately identifying "heating element" as a dryer component entity, "temperature" as an operating parameter entity, "overheating" as an abnormal operating condition type entity, and "ambient temperature" as an environmental factor entity. Relationship extraction identifies the connections between these entities, for example, discovering a causal relationship between a heating element failure and an abnormal temperature increase, and identifying anomalies in attribute values, such as temperatures outside the normal range.

[0039] Step S1163: Perform statistical analysis on the association relationships of the same entity combinations in multiple fault cases, calculate the frequency of co-occurrence of different entity combinations in fault cases, and use entity combinations and their relationships whose frequencies exceed a preset value as antecedents of candidate inference rules.

[0040] Perform statistical analysis on a large number of failure cases and calculate the frequency with which different entity combinations appear simultaneously within the failure cases. For example, count the number of instances where "heating element failure" and "abnormal temperature rise" appear simultaneously in many failure cases. When the frequency of a particular entity combination and its relationship exceeds a pre-set value, indicating that the relationship is universal and reliable, it is considered the antecedent of a candidate inference rule. This pre-set value needs to be adjusted based on actual conditions to ensure that the selected candidate inference rules have a high degree of credibility.

[0041] Step S1164: Analyze the correspondence between the candidate inference rule antecedent and the fault type diagnosis result. When the candidate inference rule antecedent appears, if the probability of a specific fault type diagnosis result exceeding a preset threshold, the fault type is used as the rule consequent to form a preliminary inference rule.

[0042] After determining the candidate inference rule antecedent, analyze its correspondence with the fault type diagnosis results. When the candidate inference rule antecedent appears, calculate the probability of a specific fault type diagnosis result. For example, when "heating element failure" and "abnormal temperature rise" occur simultaneously, calculate the probability of an "overheating" fault type diagnosis result. If this probability exceeds a pre-set threshold, indicating a strong correlation between the antecedent and the fault type, the fault type is used as the rule consequent to form a preliminary inference rule, such as "If the heating element fails and the temperature rises abnormally, the dryer will experience an abnormal overheating condition."

[0043] Step S1165: Conflict detection and duplication removal are performed on the preliminary inference rules. When the antecedents of two rules are the same but the consequents are different, the rule with more supporting cases is retained.

[0044] Preliminary inference rules may conflict or be repeated. When two rules have the same antecedent but different consequents, they conflict. To ensure the consistency and reliability of the inference rules, conflict detection and deduplication are required. Rules with a greater number of supporting cases are retained, as a greater number of supporting cases indicates that the rule is more universal and reliable. For example, there are two rules, both with the antecedent being "the heating element is faulty and the temperature is abnormally high." The consequent of one rule is "overheating," while the consequent of the other rule is "other faults." If the "overheating" rule has more supporting cases, retain that rule.

[0045] Step S1166: Assign a rule confidence to each inference rule. The rule confidence is calculated based on the ratio of the number of fault cases supporting the inference rule to the total number of cases, and finally form an inference rule set including rule antecedents, rule consequents and rule confidences.

[0046] To measure the reliability of each inference rule, a rule confidence is assigned. Rule confidence is calculated as the ratio of the number of failure cases that support the inference rule to the total number of cases. The greater the number of supporting cases, the higher the rule confidence, indicating a more reliable inference rule. For example, if a large number of failure cases support an inference rule, the rule confidence is relatively high. The resulting inference rule set consists of the rule antecedent, rule consequent, and rule confidence.

[0047] Step S120: Perform association mapping processing on the operating parameter data collected by the dryer in real time and the diagnostic knowledge graph, establish an association relationship between the real-time operating parameter data and the attribute values ​​of the corresponding entity, and generate knowledge graph instantiation data containing real-time attribute values.

[0048] Step S121: performing data cleaning on the operating parameter data collected in real time by the dryer, removing missing values ​​and abnormal fluctuation values, and retaining valid parameter data, which includes parameter name, collection timestamp and parameter value.

[0049] During the operation of the dryer, various operating parameter data are collected in real time through sensors and other equipment. However, this data may have some quality issues, such as missing values ​​and abnormal fluctuations. Missing values ​​may be caused by sensor failure, data transmission interruption, etc. These missing data will affect subsequent analysis and processing, so they need to be removed. Abnormal fluctuations refer to situations where the parameter value clearly does not conform to the normal variation pattern, which may be caused by factors such as external interference and sensor errors. For example, the temperature parameter suddenly takes on a value that is significantly different from the normal fluctuation range. Such abnormal fluctuations will mislead the judgment of the dryer's operating status, so they must also be eliminated.

[0050] During the data cleaning process, several data processing methods are employed. For missing values, interpolation can be performed based on the parameter values ​​at previous and subsequent time points, or data records containing missing values ​​can be directly deleted. For abnormal fluctuations, thresholds can be set. When a parameter value exceeds a set percentage from the normal range, it is considered an abnormal fluctuation and removed. After data cleaning, the remaining valid parameter data includes the parameter name, collection timestamp, and parameter value. The parameter name clarifies the meaning of the parameter, the collection timestamp records the time when the parameter data was collected, and the parameter value is the specific value of the parameter at the time of collection.

[0051] Step S122: traverse the operation parameter entities in the diagnosis knowledge graph, and extract the parameter name attribute value in the attribute set of each operation parameter entity.

[0052] After data cleaning is complete, the real-time collected operating parameter data needs to be matched with the operating parameter entities in the diagnostic knowledge graph. First, all operating parameter entity nodes in the diagnostic knowledge graph are traversed. In the graph database, each operating parameter entity node can be accessed through a query statement, and the parameter name attribute values ​​in its attribute set can be extracted. These parameter name attribute values ​​are the attributes previously defined for the operating parameter entities when constructing the diagnostic knowledge graph. They represent the names of the operating parameters included in the diagnostic knowledge graph, such as temperature, humidity, pressure, etc.

[0053] Step S123: Perform string matching on the parameter names in the real-time collected working condition parameter data and the parameter name attribute values ​​of the operating parameter entity. When the match is successful, establish an association relationship between the real-time working condition parameter data and the corresponding operating parameter entity.

[0054] Step S1231: After performing the same standardization processing on the parameter names in the real-time collected working condition parameter data and the parameter name attribute values ​​of the operating parameter entity, string matching is performed by combining exact matching and fuzzy matching. In exact matching, the standardized parameter name strings are directly compared to see if they are exactly the same. In fuzzy matching, the edit distance between the two strings is calculated. When the edit distance is less than the preset threshold, it is determined that the match is successful.

[0055] To improve matching accuracy, the parameter names in the real-time collected operating parameter data and the parameter name attribute values ​​of the operating parameter entity are first standardized, such as removing special symbols from the strings and unifying character case. Exact matching directly compares the standardized parameter name strings to see if they are exactly the same. If they are, the match is successful. For example, if the parameter name collected in real time is standardized to "temperature" and the parameter name attribute value of the operating parameter entity is also standardized to "temperature", then an exact match is successful. If an exact match is unsuccessful, fuzzy matching is used. Fuzzy matching determines whether a match is successful by calculating the edit distance between the two strings. The edit distance refers to the minimum number of operations required to convert one string into another by inserting, deleting, or replacing characters. When the edit distance is less than a pre-set threshold, the match is considered successful. The preset threshold setting needs to be adjusted according to actual conditions to balance matching accuracy and recall.

[0056] Step S1232: When multiple candidate operating parameter entities appear in fuzzy matching, other attribute information of the candidate operating parameter entities is extracted and matched with the corresponding information of the real-time operating parameter data for auxiliary matching, and the candidate operating parameter entity with the highest matching degree is selected as the associated object. The other attribute information includes parameter unit and sampling frequency.

[0057] During the fuzzy matching process, multiple candidate operating parameter entities may appear. In this case, it is necessary to extract other attribute information of the candidate operating parameter entity, such as parameter unit, sampling frequency, etc., and perform auxiliary matching with the corresponding information of the real-time operating parameter data. For example, if the unit of the temperature parameter collected in real time is a specific unit, then the entity with the parameter unit of the specific unit is selected as the associated object among the candidate operating parameter entities. By comprehensively considering this attribute information and selecting the candidate operating parameter entity with the highest matching degree as the associated object, the accuracy of the association can be improved.

[0058] Step S1233: Establish a mapping relationship table between the real-time operating condition parameter data and the successfully matched operating parameter entity, record the parameter data identifier, entity unique identifier and matching confidence. The matching confidence is set according to the result of exact matching or fuzzy matching. The confidence of exact matching is higher than that of fuzzy matching.

[0059] The mapping relationship table is established to record the association between real-time operating parameter data and successfully matched operating parameter entities. The mapping relationship table records the parameter data identifier, entity unique identifier, and matching confidence. The parameter data identifier is used to uniquely identify the operating parameter data collected in real time; the entity unique identifier is used to uniquely identify the operating parameter entity; the matching confidence is set based on the results of exact matching or fuzzy matching, with the confidence of exact matching being higher than that of fuzzy matching. For example, exact matching has a relatively high confidence level due to its high matching accuracy; fuzzy matching has a relatively low confidence level due to its certain uncertainty. By recording the matching confidence level, the reliability of different matching results can be evaluated in subsequent analysis.

[0060] Step S124: Use the parameter values ​​in the real-time operating parameter data as the real-time attribute values ​​of the corresponding operating parameter entity, update the attribute value field of the operating parameter entity in the diagnostic knowledge graph, and record the acquisition timestamp as the time mark of the attribute value.

[0061] After establishing an association between real-time operating parameter data and an operating parameter entity, the parameter values ​​in the real-time collected operating parameter data are used as the real-time attribute values ​​of the corresponding operating parameter entity. In the graph database, parameter values ​​are written to the attribute value field of the operating parameter entity node through an update operation. For example, if the real-time collected temperature parameter value is a certain value, this value is updated to the attribute value field of the temperature node in the diagnostic knowledge graph.

[0062] At the same time, the collection timestamp is recorded as a timestamp for the attribute value. The collection timestamp helps understand the chronological order of parameter data collection, which is crucial for analyzing parameter trends and anomalies. In a graph database, the collection timestamp can be stored as a new attribute value in the running parameter entity node, or associated with the parameter value for subsequent query and analysis.

[0063] Step S125: Instantiate the knowledge graph data after association mapping to generate knowledge graph instantiation data containing the current real-time attribute values ​​of all operating parameter entities, static attribute values ​​of dryer component entities, real-time attribute values ​​of environmental factor entities, and description attribute values ​​of abnormal working condition type entities. The entity nodes in the knowledge graph instantiation data contain the latest attribute value information.

[0064] After the mapping of real-time operating parameter data to operating parameter entities and the updating of attribute values ​​are completed, the entire knowledge graph data is instantiated. This instantiation combines the general structure and information in the knowledge graph with specific real-time data to generate knowledge graph instantiation data containing the current real-time attribute values ​​of all operating parameter entities, static attribute values ​​of dryer component entities, real-time attribute values ​​of environmental factor entities, and attribute values ​​describing abnormal operating condition types.

[0065] For the operating parameter entity, its current real-time attribute value is the latest parameter value obtained through real-time collection and updating. The static attribute values ​​of the dryer component entity are defined when the knowledge graph is constructed, such as component model, installation location, normal working range, etc. These attribute values ​​are relatively stable during the operation of the dryer. The real-time attribute values ​​of the environmental factor entity can be obtained by real-time monitoring of environmental data, such as ambient temperature, ambient humidity, etc. The descriptive attribute values ​​of the abnormal working condition type entity include fault feature description, probability of occurrence, degree of impact, etc. These attribute values ​​have been determined when the knowledge graph is constructed.

[0066] During the instantiation process, the knowledge graph's data structure is reorganized, integrating the latest attribute values ​​for each entity node. Each entity node contains the latest attribute value information, allowing the knowledge graph to accurately reflect the dryer's current operating status and environmental conditions. The instantiated knowledge graph data will serve as the foundation for subsequent multi-hop associative reasoning, further mining and analyzing abnormal dryer conditions.

[0067] Step S130: Perform multi-hop association reasoning on the knowledge graph instantiation data based on the inference rule set, retrieve the dryer component entities, environmental factor entities and abnormal operating condition type entities associated with the real-time operating parameter data, and construct an abnormal association path set containing entity nodes, relationship edges and attribute values.

[0068] Step S131: Taking the operating parameter entity associated with the real-time operating condition parameter data as the starting node, perform a breadth-first search in the knowledge graph instantiated data, and retrieve the abnormal operating condition type entity directly connected to the starting node through the representation relationship as the first-level associated entity.

[0069] After obtaining the instantiation data of the knowledge graph, multi-hop association reasoning begins. The operating parameter entity associated with the real-time operating parameter data is used as the starting node. For example, if the real-time collected temperature parameter is successfully associated with the temperature operating parameter entity in the diagnostic knowledge graph, the temperature node becomes the starting node.

[0070] Breadth-first search is a graph search algorithm that begins at a starting node and visits nodes in the graph layer by layer. Within the instantiated knowledge graph data, it searches for abnormal operating condition type entities that are directly connected to the starting node through a representational relationship. This representational relationship indicates the correspondence between abnormal changes in operating parameters and abnormal operating condition types. For example, if the temperature node is connected to the overheating abnormal operating condition type entity node through a representational relationship, the overheating node will be retrieved as a first-level associated entity.

[0071] Step S132: Starting from the first-level associated entity, continue to retrieve the dryer component entity connected to the abnormal operating condition type entity through the resulting relationship as the second-level associated entity.

[0072] After obtaining the first-level associated entities, we continue searching within the knowledge graph instantiated data, starting with these abnormal operating condition type entities. We search for dryer component entities connected to the abnormal operating condition type entities via a causes relationship. The causes relationship indicates which dryer component failure or anomaly caused the abnormal operating condition. For example, the Overheating Abnormal Operating Condition type entity node might be connected to the Heating Element Dryer Component entity node via a causes relationship, indicating that a heating element failure could cause an overheating abnormal condition. The Heating Element node is then retrieved as a second-level associated entity.

[0073] Step S133: Starting from the starting point node, the dryer component entity connected to the operation parameter entity through the association relationship is retrieved as a third-level association entity.

[0074] In addition to retrieving second-level associated entities from the first-level associated entity, the search also begins at the starting point (the operating parameter entity associated with the real-time operating parameter data) and searches for dryer component entities connected to the operating parameter entity through an association relationship. Associations represent the direct connection between dryer components and operating parameters. For example, the temperature operating parameter entity node might be connected to dryer component entity nodes such as the heating element and ventilation duct through an association relationship. This is because the operating status of the heating element affects the temperature, and the unobstructed ventilation duct also affects the temperature. Therefore, the heating element, ventilation duct, and other nodes are retrieved as third-level associated entities.

[0075] Step S134: Starting from the third-level associated entity, the environmental factor entity connected to the dryer component entity through the influence relationship is retrieved as the fourth-level associated entity.

[0076] After obtaining the third-level associated entities, we then search for environmental factor entities connected to these dryer component entities through influence relationships. Influence relationships indicate the impact of environmental factors on the dryer component's operating status and the occurrence of abnormal operating conditions. For example, the heating element dryer component entity node might be connected to the ambient temperature environmental factor entity node through an influence relationship, as high ambient temperatures increase the risk of heating element overheating. Therefore, the ambient temperature node is retrieved as a fourth-level associated entity.

[0077] Step S135: combining the starting point node, the first-level associated entity, the second-level associated entity, the third-level associated entity, the fourth-level associated entity and the relationship edges therebetween into a plurality of associated paths in a connection order.

[0078] After searching for all levels of associated entities, the starting point, first-level associated entities, second-level associated entities, third-level associated entities, fourth-level associated entities, and their relationship edges are combined in order to form multiple association paths. Each association path represents an association chain starting from the operating parameter entity, passing through the abnormal operating condition type entity, the dryer component entity, and finally to the environmental factor entity. For example, an association path might be: temperature node - representation relationship - overheating node - cause relationship - heating element node - influence relationship - ambient temperature node. This approach can discover the complex relationships between dryer operating status, abnormal operating conditions, component failures, and environmental factors.

[0079] Step S136: Perform anomaly detection on the attribute values ​​of the entity nodes in each association path. When the real-time attribute value of the operating parameter entity exceeds its normal value range, or the real-time attribute value of the environmental factor entity exceeds its impact threshold, the association path is determined to be an abnormal association path.

[0080] Step S1361: Extract the real-time attribute value of the running parameter entity in the associated path and its normal value range attribute value, and calculate the degree of deviation between the real-time attribute value and the normal value range. The degree of deviation is the ratio of the value of the real-time attribute value exceeding the upper and lower limits of the normal value range to the length of the normal value range interval.

[0081] For the operating parameter entities in the associated path, extract their real-time attribute values ​​and normal value range attribute values. The real-time attribute value is the current value of the operating parameter obtained through real-time collection, and the normal value range attribute value is the normal value range set for the operating parameter when constructing the knowledge graph. Calculate the degree of deviation between the real-time attribute value and the normal value range, specifically the ratio of the value of the real-time attribute value that exceeds the upper and lower limits of the normal value range to the length of the normal value range interval. For example, if the normal value range of temperature is a specific interval, and the temperature value collected in real time exceeds the upper limit of the interval, then the ratio of the exceeded value to the length of the interval is the degree of temperature deviation.

[0082] Step S1362: Extract the real-time attribute value of the environmental factor entity in the associated path and its impact threshold attribute value. When the real-time attribute value of the environmental factor entity is greater than the upper impact threshold or less than the lower impact threshold, calculate the abnormal deviation of the environmental factor. The abnormal deviation of the environmental factor is the ratio of the absolute value of the difference between the real-time attribute value and the nearest threshold to the length of the threshold interval.

[0083] For the environmental factor entities in the association path, their real-time attribute values ​​and impact threshold attribute values ​​are extracted. The impact threshold attribute values ​​include the upper impact threshold and the lower impact threshold. When the real-time attribute value of the environmental factor entity is greater than the upper impact threshold or less than the lower impact threshold, it indicates that the environmental factor may have an adverse effect on the operation of the dryer. The abnormal deviation of the environmental factor is calculated by the ratio of the absolute value of the difference between the real-time attribute value and the nearest threshold to the length of the threshold interval. For example, if the impact threshold of the ambient temperature is a specific interval, and the real-time collected ambient temperature value exceeds this interval, then the ratio of the absolute value of the difference with the nearest boundary value to the length of the interval is the abnormal deviation of the ambient temperature.

[0084] Step S1363: Set the operating parameter abnormality threshold and the environmental factor abnormality threshold. When the deviation degree of the operating parameter entity is greater than the operating parameter abnormality threshold, or the environmental factor abnormality deviation degree of the environmental factor entity is greater than the environmental factor abnormality threshold, mark the entity node as an abnormal node.

[0085] To determine whether the operating parameter entity and environmental factor entity are abnormal, set the operating parameter anomaly threshold and the environmental factor anomaly threshold. When the deviation of the operating parameter entity exceeds the operating parameter anomaly threshold, or the environmental factor anomaly deviation of the environmental factor entity exceeds the environmental factor anomaly threshold, the entity node is marked as an abnormal node. The settings of the operating parameter anomaly threshold and the environmental factor anomaly threshold need to be adjusted according to actual conditions to ensure accurate detection of anomalies.

[0086] Step S1364: Check whether there are abnormal nodes in the association path. When the path contains at least one abnormal node, further determine whether the relationship edge between the abnormal nodes meets the rule antecedent conditions in the inference rule set.

[0087] Check whether there are any abnormal nodes in the association path. If the path contains at least one abnormal node, further determine whether the relationship edges between the abnormal nodes meet the rule antecedent conditions in the inference rule set. The rule antecedents in the inference rule set describe the conditions that entity attributes must meet or the relationships that exist between entities. For example, if the antecedent of the inference rule is "abnormal temperature rise and heating element failure," if the association path contains an abnormal node with abnormal temperature rise and an abnormal node with heating element failure, and the relationship edges between them meet the rule antecedent conditions, then the association path may be abnormal.

[0088] Step S1365: If the relationship edge between the abnormal nodes meets the rule antecedent condition, the association path is determined to be an abnormal association path, and the attribute value information of the abnormal node and the satisfied inference rule are recorded.

[0089] When the relationship between abnormal nodes meets the rule antecedent conditions in the inference rule set, the association path is determined to be an abnormal association path. The attribute values ​​of the abnormal nodes, such as the real-time attribute values ​​of operating parameters and environmental factors, as well as the satisfied inference rules, are recorded. This information can assist in further analysis of the causes and impacts of the abnormal situation.

[0090] Step S137: Collect all abnormal association paths to form an abnormal association path set, where each abnormal association path includes a sequence of entity nodes arranged in order, a sequence of relationship edges connecting the entity nodes, and attribute value information of each entity node.

[0091] Finally, all paths identified as abnormal association paths are collected to form an abnormal association path set. Each abnormal association path consists of a sequentially arranged sequence of entity nodes, a sequence of relationship edges connecting the entity nodes, and attribute value information for each entity node. For example, the entity node sequence for an abnormal association path might be: temperature node - overheating node - heating element node - ambient temperature node; the relationship edge sequence might be: representation relationship - cause relationship - influence relationship; and the attribute value information for each entity node includes the real-time temperature value of the temperature node, the fault characteristic description of the overheating node, the component model and normal operating range of the heating element node, and the real-time ambient temperature value of the ambient temperature node. This abnormal association path set provides specific analysis targets for subsequent confidence calculation and abnormal operating condition warning. Further analysis of abnormal association paths can more accurately determine the type of abnormal operating condition of the dryer and the possible influencing factors.

[0092] Step S140: Calculate the confidence of each abnormal associated path in the abnormal associated path set, determine the comprehensive confidence of the path according to the weight value of each relationship edge in the path and the abnormal deviation of the entity attribute, and screen out the target abnormal associated paths whose comprehensive confidence exceeds the preset threshold.

[0093] Step S141: Assign a relationship weight value to each relationship edge in the abnormal association path. The relationship weight value is calculated based on the frequency of occurrence of the relationship edge in historical fault cases and the rule confidence of the corresponding inference rule. The relationship edge with higher frequency and higher rule confidence has a higher relationship weight value.

[0094] For example, step S1411: count the number of times each relationship edge in the historical fault case data appears in the abnormal association path, and use the ratio of the number to the total number of fault cases as the frequency weight of the relationship edge.

[0095] The number of times each edge in the historical fault case data appears in the abnormal association path is counted. This number reflects the importance of the edge in the historical fault case. The ratio of this number to the total number of fault cases is used as the frequency weight of the edge. For example, if the edge between "heating element failure" and "abnormal temperature rise" appears a certain number of times in many historical fault cases, the ratio of this number to the total number of cases is the frequency weight of the edge. The higher the frequency weight, the more important the edge is in the occurrence and spread of the abnormal condition.

[0096] Step S1412: Search for all inference rules containing the relationship edge from the inference rule set, extract the rule confidences of these inference rules, and calculate the average value of the rule confidences as the rule weight of the relationship edge.

[0097] Find all inference rules that include the edge in the inference rule set. The rule confidences of these inference rules reflect the reliability of the edge in the inference process. Extract the rule confidences of these inference rules and calculate their average as the rule weight of the edge. For example, if there are multiple inference rules that include a certain edge, each with its own rule confidence, average these confidences to obtain the rule weight of the edge.

[0098] Step S1413: setting a frequency weight coefficient and a rule weight coefficient, wherein both the frequency weight coefficient and the rule weight coefficient are values ​​greater than 0 and less than 1, and the sum of the two is 1.

[0099] To comprehensively consider the impact of frequency weight and rule weight on edge weights, set a frequency weight coefficient and a rule weight coefficient. Both coefficients are values ​​greater than 0 and less than 1, and their sum is 1. The frequency weight coefficient and rule weight coefficient settings need to be adjusted based on actual conditions to balance the effects of frequency weight and rule weight. For example, the values ​​of these two coefficients can be adjusted based on the importance of different edge relationships in real applications.

[0100] Step S1414: Add the product of the frequency weight of the relationship edge and the frequency weight coefficient to the product of the rule weight of the relationship edge and the rule weight coefficient to obtain the relationship weight value of the relationship edge.

[0101] The relationship weight of the edge is calculated by multiplying the frequency weight of the edge by the frequency weight coefficient and adding the product of the rule weight of the edge and the rule weight coefficient. For example, if the frequency weight of a certain edge is a specific value, the frequency weight coefficient is another specific value, the rule weight is another specific value, and the rule weight coefficient is another specific value, then the relationship weight of the edge can be calculated using the above calculation method.

[0102] Step S1415: normalize the relationship weight values ​​of all relationship edges so that the relationship weight values ​​range from 0 to 1, where 1 represents the highest weight and 0 represents the lowest weight.

[0103] To make the relationship weights comparable, normalize the relationship weights of all edges. Normalization maps the relationship weights to a range of 0 to 1, where 1 represents the highest weight and 0 represents the lowest weight. This normalization allows for a more intuitive comparison of the importance of different relationship edges.

[0104] Step S1416: Store the normalized relationship weight value in the relationship edge attribute of the diagnostic knowledge graph as the basic data for subsequent path confidence calculation.

[0105] The normalized relationship weights are stored in the relationship edge attributes of the diagnostic knowledge graph as the basis for subsequent path confidence calculations. These relationship weights can be directly used when calculating the confidence of abnormal association paths, improving calculation accuracy and efficiency.

[0106] Step S142: Calculate the product of the relationship weight values ​​of all relationship edges in the abnormal association path as the path relationship confidence.

[0107] After assigning a relationship weight to each edge, the product of the relationship weights of all edges in the abnormal correlation path is calculated to obtain the path relationship confidence. For example, for an abnormal correlation path containing multiple edges, the relationship weights of these edges are multiplied together to obtain the path relationship confidence of the path. The path relationship confidence reflects the overall reliability of each edge in the abnormal correlation path. If the relationship weights of each edge in the path are high, the path relationship confidence will also be high, indicating that the abnormal correlation reflected by the path is relatively reliable. Conversely, if the path contains edges with low relationship weights, the path relationship confidence will be affected, and the reliability of the path will be reduced accordingly.

[0108] Step S143: extracting the attribute abnormal deviations of all abnormal nodes in the abnormal association path, and calculating the average value of the abnormal deviations as the path attribute abnormality.

[0109] In addition to path relationship confidence, the abnormal deviation of entity attributes also needs to be considered. The abnormal deviation of the attributes of all abnormal nodes in the abnormal association path is extracted. For operational parameter entities, the abnormal deviation is the ratio of the number of times the real-time attribute value exceeds the upper and lower limits of the normal value range to the length of the normal value range interval. For environmental factor entities, the abnormal deviation is the ratio of the absolute value of the difference between the real-time attribute value and the most recent threshold to the length of the threshold interval.

[0110] The average of these abnormal deviations is calculated to obtain the path attribute abnormality. The path attribute abnormality reflects the degree to which the attributes of each abnormal node in the abnormal association path deviate from the normal state. The greater the abnormal deviation, the more serious the abnormality involved in the path.

[0111] Step S144: setting the relationship weight coefficient and the attribute anomaly coefficient, adding the product of the path relationship confidence and the relationship weight coefficient to the product of the path attribute anomaly and the attribute anomaly coefficient to obtain the path comprehensive confidence.

[0112] To comprehensively consider the impact of path relationship confidence and path attribute anomaly on path reliability, we set a relationship weight coefficient and an attribute anomaly coefficient. These coefficients adjust the weights of path relationship confidence and path attribute anomaly in the calculation of the overall path confidence. For example, the relationship weight coefficient can be set to a value that reflects the importance of relationship edges, indicating a greater emphasis on the reliability of relationship edges in the path; the attribute anomaly coefficient can be set to a value that reflects the importance of attribute anomalies, indicating the degree of emphasis on abnormal deviations of entity attributes.

[0113] The path's comprehensive confidence is calculated by multiplying the path's relationship confidence by the relationship weight coefficient and adding the path's attribute anomaly by the attribute anomaly coefficient. This comprehensive metric considers both the reliability of each edge in the anomalous association path and the degree of attribute anomaly at each anomalous node, providing a more comprehensive assessment of the credibility of the anomalous association path.

[0114] Step S145: sort all abnormal correlation paths in the abnormal correlation path set from high to low according to the path comprehensive confidence.

[0115] After calculating the comprehensive confidence level for each abnormal correlation path, all abnormal correlation paths in the abnormal correlation path set are sorted from high to low based on their comprehensive confidence level. This sorting process facilitates the subsequent screening of abnormal correlation paths with higher comprehensive confidence levels, as these paths are more likely to reflect the actual abnormality of the dryer. A sorting algorithm can be used to sort the abnormal correlation paths, placing paths with higher comprehensive confidence levels at the top and paths with lower comprehensive confidence levels at the bottom.

[0116] Step S146: setting a path confidence threshold, and screening out abnormal associated paths whose comprehensive path confidence is higher than the path confidence threshold as target abnormal associated paths.

[0117] A path confidence threshold is set to determine whether the abnormal correlation path has sufficient credibility. When the comprehensive path confidence of the abnormal correlation path is higher than the path confidence threshold, the path is screened out as the target abnormal correlation path. The target abnormal correlation path is the path that is most likely to reflect the actual abnormal operating condition of the dryer. Abnormal operating condition warning information will be generated based on these paths. The setting of the path confidence threshold can be adjusted according to the actual situation. If you want to more strictly screen out reliable abnormal correlation paths, you can set the threshold higher; if you want to obtain more abnormal correlation paths for further analysis, you can set the threshold lower.

[0118] Step S147: When the number of target abnormality associated paths exceeds a preset number, a preset number of target abnormality associated paths with a higher ranking are retained.

[0119] If the number of selected target anomaly-related paths exceeds a preset number, only the top-ranked preset number of target anomaly-related paths will be retained to avoid excessive information interference. For example, if a certain number of paths is set, and if the number of selected target anomaly-related paths exceeds this number, only those with the highest overall confidence ranking will be retained. This ensures that the focus is on the paths most likely to experience anomalies, improving the accuracy and effectiveness of early warnings.

[0120] Step S150: generating dryer abnormal operating condition warning information according to the abnormal operating condition type entity, dryer component entity and abnormal propagation relationship included in the target abnormality association path, and sending the abnormal operating condition warning information to the dryer monitoring terminal.

[0121] Step S151: extracting an abnormal operating condition type entity from the target abnormality association path, and obtaining an abnormal operating condition type identifier and a fault feature description attribute value of the abnormal operating condition type entity.

[0122] After filtering out the target abnormality association path, start generating the dryer abnormal condition warning information. First, extract the abnormal condition type entity from the target abnormality association path. For example, in a target abnormality association path, there is an overheating abnormal condition type entity node. Get the abnormal condition type identifier and fault feature description attribute value of the abnormal condition type entity. The abnormal condition type identifier is used to uniquely identify different abnormal condition types, which facilitates the monitoring terminal to classify and manage different types of abnormal conditions. The fault feature description attribute value describes in detail the specific manifestations of the abnormal condition. For example, the fault feature description of the overheating abnormal condition may be that the surface temperature of the dryer is too high, there is an odor, etc. This information can help operators quickly identify abnormal conditions.

[0123] Step S152: extract all dryer component entities included in the target abnormality association path, obtain the component name and installation location attribute value of each dryer component entity as the key influencing dryer component entity name.

[0124] At the same time, all dryer component entities are extracted from the target anomaly association path. For example, the target anomaly association path may include dryer component entity nodes such as heating elements and ventilation ducts. The component name and installation location attribute values ​​are obtained for each dryer component entity. The component name is used to identify the specific dryer component, and the installation location attribute value helps operators quickly locate the faulty component. This component name and installation location information will serve as the key influencing dryer component entity name, used to identify the specific components and locations that may be affected by the abnormal operating condition.

[0125] Step S153: Generate an exception propagation path description according to the entity node order and relationship edge order in the target exception association path, wherein the exception propagation path description includes the entity node name, relationship edge name and connection order.

[0126] An anomaly propagation path description is generated based on the order of entity nodes and relationship edges in the target anomaly association path. This description details how the abnormal operating condition begins with an abnormal change in operating parameters and gradually propagates and develops through the association relationships between different entities. For example, the entity node order of a target anomaly association path is: temperature node - overheating node - heating element node - ambient temperature node, and the relationship edge order is: representation relationship - cause relationship - influence relationship. The anomaly propagation path description could then be: the abnormal change in temperature parameters triggers an abnormal overheating condition through the representation relationship, which is then linked to a heating element failure through the cause relationship. Meanwhile, the ambient temperature affects the entire abnormal situation through the influence relationship. This description, which includes entity node names, relationship edge names, and connection order, allows operators to clearly understand the occurrence and development of the abnormal condition, enabling them to better formulate response strategies.

[0127] Step S154: Combine the abnormal operating condition type identifier, the entity name of the key dryer-affecting component, and the abnormal propagation path description into structured warning information content, and add a timestamp to the warning information content. The timestamp is the current system time when the warning information is generated.

[0128] The abnormal operating condition type identifier, the entity name of the key dryer component that affects the condition, and the description of the abnormal propagation path are combined into structured warning information content. Structured warning information content is easy for operators to quickly understand and process. For example, this information can be organized according to a set format, such as the abnormal operating condition type identifier first, the entity name of the key dryer component that affects the condition second, and the description of the abnormal propagation path last. At the same time, a timestamp is added to the warning information content, which is the current system time when the warning information is generated. The timestamp can record the generation time of the warning information, facilitating subsequent tracing and analysis. Operators can understand the order and duration of abnormal operating conditions based on the timestamp.

[0129] Step S155: Determine the warning level corresponding to the comprehensive path confidence of the target abnormality associated path according to the pre-built mapping relationship table.

[0130] Based on a pre-built mapping table, the warning level corresponding to the comprehensive confidence level of the target anomaly-associated path is determined. This mapping table specifies the warning levels corresponding to different ranges of comprehensive confidence levels. Warning levels can be divided into different grades to indicate the severity of the abnormal operating condition. For example, when the comprehensive confidence level of the target anomaly-associated path is in a high range, the corresponding warning level is high; when the comprehensive confidence level is in a medium range, the corresponding warning level is medium; and when the comprehensive confidence level is in a low range, the corresponding warning level is low. By determining the warning level, operators can quickly understand the severity of the abnormal situation and take appropriate measures.

[0131] Step S156: Encapsulate the warning information content, timestamp and warning level into dryer abnormal operating condition warning information.

[0132] Finally, the warning information content, timestamp, and warning level are packaged into a dryer abnormal operating condition warning message. The packaged warning information contains detailed information about the abnormal operating condition, the time it was generated, and the severity, and is a complete warning data packet. The warning information can be packaged in a set data format to facilitate transmission and processing within the network. This warning information is sent to the dryer monitoring terminal, which can display the warning information to the operator so that they can take timely measures to resolve the dryer's abnormality and ensure the normal operation of the dryer. The operator can develop a corresponding maintenance and treatment plan based on the abnormal operating condition type, key influencing components, abnormal propagation path, and warning level in the warning information to ensure that the dryer can resume normal operation as soon as possible.

[0133] In summary, this embodiment achieves accurate analysis and timely warning of dryer abnormal conditions by constructing a diagnostic knowledge graph for dryer abnormal conditions, associating and mapping real-time collected operating parameter data with the knowledge graph, performing multi-hop association reasoning, calculating the confidence of abnormal association paths, screening target abnormal association paths, and generating and transmitting abnormal condition warning information based on the target abnormal association paths. This entire method process fully utilizes historical and real-time dryer data, combined with the structure and inference rules of the knowledge graph, to effectively identify abnormal dryer conditions and provide detailed warning information to operators, helping to improve the operational reliability and safety of the dryer.

[0134] Further, Figure 2 FIG. 1 shows a hardware structure diagram of a dryer abnormal operating condition data analysis and early warning system 100 for implementing the method provided in an embodiment of the present application. Figure 2 As shown, the abnormal operating condition data analysis and early warning system 100 for the dryer may include at least one processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, a transmission device 106 for communication functions, and a controller 108. It will be understood by those skilled in the art that Figure 2 The structure shown is only for illustration and does not limit the structure of the abnormal working condition data analysis and early warning system 100 for the dryer. For example, the abnormal working condition data analysis and early warning system 100 for the dryer may also include Figure 2 More or fewer components than shown, or with Figure 2 Different configurations shown.

[0135] The memory 104 can be used to store software programs and modules of application software, such as the program instructions corresponding to the method embodiments described above in the embodiments of the present application. The processor 102 executes the software programs and modules stored in the memory 104 to perform various functional applications and data processing, thereby implementing the above-mentioned abnormal operating condition data analysis and early warning method for the dryer. The transmission device 106 is used to obtain or send data via a network.

[0136] Those skilled in the art will understand that all or part of the steps for implementing the above embodiments may be accomplished by hardware, or may be accomplished by instructing the relevant hardware through a program, and the above program may be stored in a computer-readable storage medium, which may be a read-only memory, a disk, or an optical disk, etc.

Claims

1. A method for analyzing and warning abnormal operating data of a dryer, characterized in that: The method comprises: Construct a diagnostic knowledge graph for abnormal operating conditions of a dryer. The diagnostic knowledge graph includes an entity type set, a relationship type set, an attribute set, and an inference rule set. The entity type set includes dryer component entities, operating parameter entities, abnormal operating condition type entities, and environmental factor entities. The relationship type set includes the association relationship between components and parameters, the representation relationship between parameters and operating conditions, and the impact relationship between operating conditions and the environment. The attribute set includes feature description information corresponding to each entity type. The inference rule set includes abnormality propagation logic rules based on the relationship between entities. Performing an association mapping process on the operating parameter data collected in real time by the dryer and the diagnostic knowledge graph, establishing an association relationship between the real-time operating parameter data and the attribute values ​​of the corresponding entities, and generating knowledge graph instantiation data containing the real-time attribute values; Based on the inference rule set, multi-hop association reasoning is performed on the instantiated data of the knowledge graph, and the dryer component entity, environmental factor entity and abnormal operating condition type entity associated with the real-time operating condition parameter data are retrieved, and an abnormal association path set including entity nodes, relationship edges and attribute values ​​is constructed; Calculate the confidence of each abnormal associated path in the abnormal associated path set, determine the comprehensive confidence of the path according to the weight value of each relationship edge in the path and the abnormal deviation of the entity attribute, and screen out the target abnormal associated paths whose comprehensive confidence exceeds a preset threshold; Dryer abnormal operating condition warning information is generated according to the abnormal operating condition type entity, dryer component entity and abnormal propagation relationship included in the target abnormality association path, and the abnormal operating condition warning information is sent to the dryer monitoring terminal.

2. The abnormal operating condition data analysis and early warning method for the dryer according to claim 1 is characterized in that: The construction of the diagnostic knowledge graph of the dryer abnormal operating condition includes: Collect dryer design documents, operation and maintenance manuals, and historical failure case data, and extract dryer component names, operating parameter names, abnormal operating condition type names, and environmental factor names as entity candidate sets; Performing entity type labeling on the entity candidate set, dividing the entity candidate set into dryer component entities, operating parameter entities, abnormal operating condition type entities, and environmental factor entities to form an entity type set; Based on the historical fault case data and operation and maintenance manual, the association relationships between different entity types are analyzed, and the association relationships between components and parameters, the representation relationships between parameters and working conditions, the impact relationships between working conditions and the environment, and the causal relationships between components and working conditions are defined to form a relationship type set; Define a set of attributes for each entity type. The attributes of the dryer component entity include component model, installation location, and normal operating range; the attributes of the operating parameter entity include parameter unit, sampling frequency, and normal value range; the attributes of the abnormal operating condition type entity include fault feature description, occurrence probability, and impact level; the attributes of the environmental factor entity include factor category, measurement range, and impact threshold; Based on the entity type set, relationship type set and attribute set, a graph database is used to construct the basic structure of the diagnostic knowledge graph, wherein entities are nodes of the graph, relationships are edges connecting nodes, and attributes are stored as attribute values ​​of nodes; The abnormal propagation logic is extracted from historical fault case data and converted into a set of inference rules of the diagnostic knowledge graph. The inference rule set includes rule antecedents and rule consequents. The rule antecedents are the conditions satisfied by entity attributes or the relationships between entities, and the rule consequents are the new relationships obtained by reasoning or the judgment results of abnormal working condition types.

3. The abnormal operating condition data analysis and early warning method for the dryer according to claim 2 is characterized in that: The method extracts the anomaly propagation logic from the historical fault case data and converts it into a set of inference rules for the diagnostic knowledge graph, including: Structural processing of historical fault case data. Each fault case includes the time of fault occurrence, a list of involved components, a list of abnormal parameters, environmental factor records, and fault type diagnosis results. Natural language processing technology is used to perform entity recognition and relationship extraction on the text description of the fault case, identifying the dryer component entities, operating parameter entities, abnormal operating condition type entities, and environmental factor entities contained in the case, and extracting the association relationships between entities and abnormal attribute values; Perform statistical analysis on the association relationships of the same entity combinations in multiple fault cases, calculate the frequency of different entity combinations appearing in fault cases, and select entity combinations and their relationships with frequencies exceeding a preset value as the antecedents of candidate inference rules; Analyze the correspondence between the candidate inference rule antecedents and the fault type diagnosis results. When the candidate inference rule antecedents appear, if the probability of a specific fault type diagnosis result exceeding a preset threshold, then use the fault type as the rule consequent to form a preliminary inference rule. Conflict detection and duplicate removal are performed on the preliminary inference rules. When two rules have the same antecedents but different consequents, the rule with more supporting cases is retained. A rule confidence is assigned to each inference rule. The rule confidence is calculated based on the ratio of the number of failure cases supporting the inference rule to the total number of cases, and finally an inference rule set including rule antecedents, rule consequents and rule confidence is formed.

4. The abnormal operating condition data analysis and early warning method for a dryer according to claim 1 is characterized in that: The process of performing an association mapping process on the operating parameter data collected in real time by the dryer and the diagnostic knowledge graph, establishing an association relationship between the real-time operating parameter data and the attribute values ​​of the corresponding entities, and generating knowledge graph instantiation data containing the real-time attribute values ​​includes: Perform data cleaning on the operating parameter data collected in real time by the dryer to remove missing values ​​and abnormal fluctuation values ​​and retain valid parameter data, which includes parameter name, collection timestamp and parameter value; Traversing the operation parameter entities in the diagnostic knowledge graph, and extracting the parameter name attribute value in the attribute set of each operation parameter entity; Perform string matching on the parameter names in the real-time collected working condition parameter data and the parameter name attribute values ​​of the operating parameter entity. When the match is successful, establish an association relationship between the real-time working condition parameter data and the corresponding operating parameter entity; Use the parameter values ​​in the real-time operating parameter data as the real-time attribute values ​​of the corresponding operating parameter entity, update the attribute value field of the operating parameter entity in the diagnostic knowledge graph, and record the acquisition timestamp as the time stamp of the attribute value; The knowledge graph data after association mapping is instantiated to generate knowledge graph instantiation data containing the current real-time attribute values ​​of all operating parameter entities, static attribute values ​​of dryer component entities, real-time attribute values ​​of environmental factor entities, and description attribute values ​​of abnormal working condition type entities. The entity nodes in the knowledge graph instantiation data contain the latest attribute value information.

5. The abnormal operating condition data analysis and early warning method for a dryer according to claim 4 is characterized in that: The parameter name in the real-time collected working condition parameter data is matched with the parameter name attribute value of the operating parameter entity, and when the match is successful, an association relationship between the real-time working condition parameter data and the corresponding operating parameter entity is established, including: After the parameter names in the real-time collected working condition parameter data and the parameter name attribute values ​​of the operating parameter entity are subjected to the same standardization processing, string matching is performed by combining exact matching and fuzzy matching. In exact matching, the standardized parameter name strings are directly compared to see if they are exactly the same. In fuzzy matching, the edit distance between the two strings is calculated. When the edit distance is less than a preset threshold, it is determined that the match is successful. When multiple candidate operating parameter entities appear in the fuzzy matching, other attribute information of the candidate operating parameter entities is extracted and matched with the corresponding information of the real-time operating parameter data for auxiliary matching, and the candidate operating parameter entity with the highest matching degree is selected as the associated object. The other attribute information includes parameter unit and sampling frequency. A mapping relationship table is established between real-time operating condition parameter data and successfully matched operating parameter entities, recording the parameter data identifier, entity unique identifier and matching confidence. The matching confidence is set according to the results of exact matching or fuzzy matching, and the confidence of exact matching is higher than that of fuzzy matching.

6. The abnormal operating condition data analysis and early warning method for a dryer according to claim 1 is characterized in that: The method performs multi-hop association reasoning on the instantiated data of the knowledge graph based on the inference rule set, retrieves dryer component entities, environmental factor entities, and abnormal operating condition type entities associated with the real-time operating condition parameter data, and constructs an abnormal association path set containing entity nodes, relationship edges, and attribute values, including: Taking the operating parameter entity associated with the real-time operating parameter data as the starting node, a breadth-first search is performed in the instantiated data of the knowledge graph to retrieve the abnormal operating condition type entity directly connected to the starting node through the representation relationship as the first-level associated entity; Starting from the first-level associated entity, continue to retrieve the dryer component entity connected to the abnormal working condition type entity through the resulting relationship as the second-level associated entity; At the same time, starting from the starting point node, the dryer component entity connected to the operation parameter entity through the association relationship is retrieved as the third-level association entity; Starting from the third-level associated entity, retrieve the environmental factor entity connected to the dryer component entity through the influence relationship as the fourth-level associated entity; Combining the starting point node, the first-level associated entity, the second-level associated entity, the third-level associated entity, the fourth-level associated entity and the relationship edges therebetween into multiple associated paths in a connection order; Perform anomaly detection on the attribute values ​​of the entity nodes in each association path. When the real-time attribute value of the operating parameter entity exceeds its normal value range, or the real-time attribute value of the environmental factor entity exceeds its impact threshold, the association path is determined to be an abnormal association path. All abnormal association paths are collected to form an abnormal association path set. Each abnormal association path contains a sequence of entity nodes arranged in order, a sequence of relationship edges connecting the entity nodes, and attribute value information of each entity node.

7. The abnormal operating condition data analysis and early warning method for a dryer according to claim 6, characterized in that: The abnormality detection is performed on the attribute values ​​of the entity nodes in each association path. When the real-time attribute value of the operation parameter entity exceeds its normal value range, or the real-time attribute value of the environmental factor entity exceeds its impact threshold, the association path is determined to be an abnormal association path, including: Extract the real-time attribute values ​​and their normal value range attribute values ​​of the operation parameter entities in the associated path, and calculate the degree of deviation between the real-time attribute values ​​and the normal value range. The degree of deviation is the ratio of the value of the real-time attribute value exceeding the upper and lower limits of the normal value range to the length of the normal value range interval; Extract the real-time attribute value and its impact threshold attribute value of the environmental factor entity in the association path. When the real-time attribute value of the environmental factor entity is greater than the upper impact threshold or less than the lower impact threshold, calculate the environmental factor abnormal deviation degree. The environmental factor abnormal deviation degree is the ratio of the absolute value of the difference between the real-time attribute value and the nearest threshold to the length of the threshold interval. Set an abnormal threshold for operating parameters and an abnormal threshold for environmental factors. When the deviation degree of the operating parameter entity is greater than the abnormal threshold for operating parameters, or the abnormal deviation degree of the environmental factor entity is greater than the abnormal threshold for environmental factors, mark the entity node as an abnormal node. Check whether there are abnormal nodes in the association path. If the path contains at least one abnormal node, further determine whether the relationship edge between the abnormal nodes meets the rule antecedent conditions in the inference rule set; If the relationship edge between abnormal nodes meets the rule antecedent condition, the association path is determined to be an abnormal association path, and the attribute value information of the abnormal node and the satisfied inference rule are recorded.

8. The abnormal operating condition data analysis and early warning method for a dryer according to claim 1 is characterized in that: The confidence calculation is performed on each abnormal associated path in the abnormal associated path set, the comprehensive confidence of the path is determined according to the weight value of each relationship edge in the path and the abnormal deviation of the entity attribute, and the target abnormal associated path with a comprehensive confidence exceeding a preset threshold is screened out, including: Assign a relationship weight value to each relationship edge in the abnormal association path. The relationship weight value is calculated based on the frequency of the relationship edge in historical fault cases and the rule confidence of the corresponding inference rule. The relationship edge with higher frequency and higher rule confidence has a higher relationship weight value; Calculate the product of the relationship weight values ​​of all relationship edges in the abnormal association path as the path relationship confidence; Extract the attribute abnormal deviation of all abnormal nodes in the abnormal association path, and calculate the average value of each abnormal deviation as the path attribute abnormality; Set the relationship weight coefficient and attribute anomaly coefficient, add the product of the path relationship confidence and the relationship weight coefficient to the product of the path attribute anomaly and the attribute anomaly coefficient to obtain the path comprehensive confidence; Sort all abnormal correlation paths in the abnormal correlation path set from high to low according to the comprehensive confidence of the paths; Set a path confidence threshold and select abnormal associated paths whose comprehensive path confidence is higher than the path confidence threshold as target abnormal associated paths; When the number of target abnormality association paths exceeds a preset number, the preset number of target abnormality association paths with the highest ranking are retained.

9. The abnormal operating condition data analysis and early warning method for a dryer according to claim 1, characterized in that: The generating of the dryer abnormal operating condition warning information according to the abnormal operating condition type entity, the dryer component entity and the abnormal propagation relationship included in the target abnormal association path includes: Extract the abnormal working condition type entity from the target abnormal association path, and obtain the abnormal working condition type identifier and fault feature description attribute value of the abnormal working condition type entity; Extract all dryer component entities contained in the target anomaly association path, obtain the component name and installation location attribute value of each dryer component entity as the key influencing dryer component entity name; Generate an exception propagation path description according to the entity node order and relationship edge order in the target exception association path, wherein the exception propagation path description includes the entity node name, relationship edge name and connection order; Combine the abnormal operating condition type identifier, the entity name of the key dryer component affecting the abnormality, and the abnormal propagation path description into structured warning information content, and add a timestamp to the warning information content, where the timestamp is the current system time when the warning information is generated; Determine the warning level corresponding to the comprehensive confidence level of the target abnormal associated path according to the pre-built mapping relationship table; The warning information content, timestamp and warning level are encapsulated into the dryer abnormal working condition warning information.

10. A data analysis and early warning system for abnormal operating conditions of a dryer, characterized in that: The device comprises a processor and a readable storage medium, wherein the readable storage medium stores a program, and when the program is executed by the processor, the method for analyzing and warning abnormal operating condition data of a dryer according to any one of claims 1 to 9 is implemented.

Images