Industrial Internet equipment fault diagnosis method and system

Abstract

The present invention discloses a method and system for diagnosing faults in industrial Internet equipment, specifically relating to the field of fault-tolerant diagnosis of industrial Internet equipment. The method comprises inputting the original control instruction sequence from the industrial equipment control bus into a structured parser to perform syntax classification and function attribution extraction, and outputting a path set with logical uniqueness; inputting the path set into a path label engine to perform structural classification mapping, and outputting a redundant logical path set including a primary execution path, a secondary execution path, and a mirror verification path; and inputting the redundant logical path set and a standard operating condition sequence into an execution environment. By establishing a dynamic mapping relationship between the semantic trajectory of the redundant path and the input stream, and continuously tracking its semantic evolution trend, early identification of redundant logical paths with "normal output results on the surface but drifted semantics" is achieved, solving the problem of silent data corruption that cannot be detected due to the failure of the redundancy mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of fault tolerance diagnosis of industrial Internet equipment. More specifically, the present invention relates to a method and system for fault diagnosis of industrial Internet equipment. Background Art

[0002] In cloud-edge collaborative environments, to enhance the reliability of critical computing results, systems often introduce redundant logic paths for parallel computing and result verification. However, as the system runs over time, these redundant paths may gradually deviate from their logical behavior due to micro-evolutions such as operational optimization and floating-point carry drift.

[0003] In this case, although the output result is numerically consistent with the target, its semantics has deviated from the original logical intention, resulting in "silent data corruption";

[0004] Since the current verification mechanism assumes that redundant paths are trusted sources by default, it makes it difficult for the system to identify such errors and mistakenly accept them as valid results, eventually evolving into highly hidden faults that are imperceptible and difficult to locate. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an industrial Internet equipment fault diagnosis method and system. By establishing a dynamic mapping relationship between the redundant path semantic trajectory and the input stream, and continuously tracking its semantic evolution trend, it realizes the early identification of redundant logical paths where "the output results appear normal but the semantics have drifted", so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a method for diagnosing faults in industrial Internet equipment, comprising:

[0007] The original control instruction sequence from the industrial equipment control bus is input into the structured parser to perform syntax classification and function attribution extraction, and output a set of logically unique paths;

[0008] Input the path set into the path label engine to perform structure classification mapping, and output a redundant logical path set including the main execution path, the secondary execution path, and the image verification path;

[0009] The redundant logic path set and the standard working condition sequence are input into the execution environment, the fixed input operation process is executed, and the path output data corresponding to the path is output;

[0010] The path output data sequence is input into the trend modeling mechanism to perform continuity analysis and boundary identification, and output a semantic response profile with boundary constraint characteristics;

[0011] The input-output pairs consisting of the standard working condition sequence and the path output data sequence are input into the semantic projection builder together with the semantic response profile, and semantic mapping modeling is performed to output the semantic behavior map.

[0012] Input the semantic behavior graph into the trust initialization model to perform consistency comparison between paths and output the initial path trust weight graph;

[0013] By mapping input flows to redundant paths and tracing their temporal semantic trajectories, a structural surface is formed to express the evolutionary state of trust between paths, providing dynamic input for subsequent risk modeling and risk assessment, thereby establishing a quantifiable evolutionary foundation for fault state identification and fault-tolerant mechanism response.

[0014] In a preferred embodiment, the control input stream collected during the operation of the device is mapped to a set of redundant logical paths, real-time functional operations are performed and the results of the path operation are collected, and a sequence of output values ​​of the path operation is output;

[0015] The runtime output value sequence of the path is structurally aligned with the semantic behavior graph, and a dynamic semantic projection modeling process is executed to output a semantic trajectory point set during the runtime of the path. The semantic trajectory point set is divided into fixed time windows, and local aggregation and structured segmentation processing are performed to output a trajectory segment sequence containing continuous segments, mutation segments, and drift segments.

[0016] Perform disturbance structure identification and trust change trend modeling on the trajectory segment sequence, and output the path trust attenuation factor and disturbance entropy value set;

[0017] The path trust attenuation factor and the disturbance entropy value set are used to perform cross-path fusion modeling to construct a redundant path trust evolution structure surface to express the trust evolution status between paths, and form a dynamic input basis for risk modeling and risk assessment, supporting fault state identification and quantification of fault tolerance mechanism response.

[0018] In a preferred embodiment, the path trust evolution structure surface is used as input, and a pattern recognition and variation density statistical process is performed to output a set of risk factors that characterize path stability;

[0019] The risk factor set is input into the directional consistency analysis process, the intersection of drift directions between paths is extracted, and the path resonance drift intersection point set is output;

[0020] The resonance drift intersection set is input into the risk expression construction process, the graph generation operation is performed, and the trust gradient expression graph is output; the trust gradient expression graph is structurally fused with the current path execution graph, and a diagnostic starting point layer with response trigger characteristics is output.

[0021] In a preferred embodiment, the diagnostic response starting point layer is used as input, the path space position mapping and trust weight threshold state identification process is executed, and the set of paths to be stripped containing path identification, function type and state label is output;

[0022] Taking the set of paths to be stripped as input, the operations of disconnecting the master connection, canceling the path fusion permission, and clearing the status cache are performed in sequence. The structural position, connection dependency, current status, and cancellation feedback information of each path are extracted to generate a path stripping status information table. The spatial location distribution, functional type classification, and stripping mark status recorded in the stripping status information table are used to construct a path resource distribution map in the stripping state.

[0023] In a preferred embodiment, the path resource distribution map in the stripped state is used as input. Based on the functional label, semantic behavior characteristics, and historical response sequence of each path to be replaced, a consistency screening process based on semantic mapping relationships is performed to output a set of candidate replacement paths that meet functional equivalence, semantic projection overlap, and behavioral convergence stability.

[0024] The candidate replacement path set is subjected to semantic fitting evaluation and boundary consistency verification. The semantic fitting evaluation and boundary consistency verification include path semantic vector reconstruction, behavior boundary response rate calculation and fusion compatibility verification, and the adaptation path that meets the replacement conditions is output.

[0025] In a preferred embodiment, a virtual load response test is performed on the adaptation path. The virtual load response test monitors its behavior convergence and response stability based on the simulated input flow required by the current master control scheduling, and outputs a verification pass path that meets the convergence conditions;

[0026] Connect the verified path to the main path execution structure, synchronize the execution path structure replacement with the result, register the path replacement time point, alternative path identifier and evolution trust value to the trust evolution record system, and output the updated path execution structure diagram.

[0027] In a preferred embodiment, the trust trajectory of the stripped path before replacement, the corresponding risk factor, and the verified replacement path information are used as input, and the path behavior archiving, label standardization, and evolution stage marking processes are performed to output a structurally complete degradation learning sample set;

[0028] The trust trajectory of the current running path is aligned with the degradation learning sample set for feature vector analysis, trend direction fitting and error interval calculation are performed, and the degradation trend label of the current path is output.

[0029] In a preferred embodiment, the degradation trend label of the current path is input into the path trust intervention mechanism, trust limit adjustment, access permission reduction and alternative path scheduling initialization operations are performed, and the corresponding path trust control instruction set is output;

[0030] Taking the path trust control instruction set as input, the recording and archiving process of the path behavior intervention results is executed. The recording and archiving process includes control response data, path switching process and post-intervention status identification, and outputs a knowledge graph for updating the path evolution logical structure.

[0031] An industrial Internet equipment fault diagnosis system includes a path extraction module, a redundant path module, a calculation module, a modeling semantic boundary module, a behavior graph module, a trust relationship module, and a trust structure surface module;

[0032] The path extraction module is used to input the original control instruction sequence from the industrial equipment control bus into the structured parser to perform syntax classification and function attribution extraction, and output a path set with logical uniqueness;

[0033] The redundant path module is used to input the path set into the path label engine to perform structure classification mapping, and output a redundant logical path set including the main execution path, the secondary execution path, and the mirror verification path;

[0034] The operation module is used to input the redundant logic path set and the standard working condition sequence into the execution environment, execute the fixed input operation process, and output the path output data corresponding to the path;

[0035] The modeling semantic boundary module is used to input the path output data sequence into the trend modeling mechanism to perform continuity analysis and boundary identification, and output a semantic response profile with boundary constraint characteristics;

[0036] The behavior graph module is used to input the input-output pairs consisting of the standard working condition sequence and the path output data sequence, and input them together with the semantic response profile into the semantic projection builder, perform semantic mapping modeling, and output the semantic behavior graph;

[0037] The trust relationship module is used to input the semantic behavior graph into the trust initialization model to perform consistency comparison between paths and output the initial path trust weight graph;

[0038] The trust structure plane module maps input flows to redundant paths and tracks their temporal semantic trajectories to form a structural plane that expresses the evolutionary state of trust between paths, providing dynamic input for subsequent risk modeling and risk assessment, thereby establishing a quantifiable evolutionary foundation for fault state identification and fault-tolerant mechanism response.

[0039] Technical effects and advantages of the present invention:

[0040] 1. By establishing a dynamic mapping relationship between redundant path semantic traces and input streams and continuously tracking their semantic evolution trends, this system enables early identification of redundant logical paths where the output appears normal but the semantics have drifted. This solves the problem of silent data corruption that cannot be detected due to redundancy mechanism failure.

[0041] 2. By constructing multidimensional perturbation functions such as perturbation velocity, perturbation acceleration, and directional frequency, and integrating them with perturbation complexity functions and perturbation entropy values, we achieve dynamic quantification of path operation stability and trend degradation, providing a nonlinear perception foundation for subsequent trust evolution modeling and risk assessment.

[0042] 3. By constructing a semantic behavior graph and processing a trust initialization model, we achieve path-level semantic boundary modeling and inter-path consistency quantification. This allows us to establish a multi-path trust relationship structure during the path initialization phase, providing structural constraints for subsequent fault-tolerance strategy formulation.

[0043] 4. By constructing a path behavior difference intensity function and introducing a nested calculation logic combining semantic direction angle, boundary response rate, and semantic feature difference, we implemented a mechanism for scoring and ranking the suitability of candidate alternative paths, improving the consistency of alternative paths in both behavioral and structural dimensions.

[0044] 5. By establishing a fusion processing mechanism between the diagnostic response starting point layer and the path execution graph, the system can pre-position potential high-risk path structures during operation, improving the system's proactive perception and response capabilities to structural hazards.

[0045] 6. By constructing a degradation learning sample set and aligning it with the path trust trajectory for analysis, and introducing nonlinear scoring control instruction generation logic, a closed loop of intervention and evolution of the path life cycle is established, enabling the system to perceive, autonomously identify, and dynamically correct path degradation behavior in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The figure is a flow chart of the method steps of the present invention.

[0047] Figure 2 It is a system module diagram of the present invention.

[0048] Figure 3 Initializes the control path flow chart of the present invention.

[0049] Figure 4 This is a flowchart of the runtime trust evolution of the present invention.

[0050] Figure 5 A flow chart is generated for the risk response initiation layer of the present invention.

[0051] Figure 6This is a flow chart of the path stripping and replacement mechanism of the present invention.

[0052] Figure 7 Archive flow chart for the path intervention and evolution of the present invention. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0054] Refer to the instruction manual Figure 1-7 , a method for diagnosing faults of industrial Internet equipment according to an embodiment of the present invention includes:

[0055] The original control instruction sequence from the industrial equipment control bus is input into the structured parser to perform syntax classification and function attribution extraction, and output a set of logically unique paths;

[0056] Input the path set into the path label engine to perform structure classification mapping, and output a redundant logical path set including the main execution path, the secondary execution path, and the image verification path;

[0057] The redundant logic path set and the standard working condition sequence are input into the execution environment, the fixed input operation process is executed, and the path output data corresponding to the path is output;

[0058] The path output data sequence is input into the trend modeling mechanism to perform continuity analysis and boundary identification, and output a semantic response profile with boundary constraint characteristics;

[0059] The input-output pairs consisting of the standard working condition sequence and the path output data sequence are input into the semantic projection builder together with the semantic response profile, and semantic mapping modeling is performed to output the semantic behavior map.

[0060] Input the semantic behavior graph into the trust initialization model to perform consistency comparison between paths and output the initial path trust weight graph;

[0061] By mapping input flows to redundant paths and tracing their temporal semantic trajectories, a structural surface is formed to express the evolutionary state of trust between paths, providing dynamic input for subsequent risk modeling and risk assessment, thereby establishing a quantifiable evolutionary foundation for fault state identification and fault-tolerant mechanism response.

[0062] Map the control input stream collected during the equipment operation period to a redundant logic path set, perform real-time functional operations and collect path operation results, and output the output value sequence of the path operation period;

[0063] The runtime output value sequence of the path is structurally aligned with the semantic behavior graph, and a dynamic semantic projection modeling process is executed to output a semantic trajectory point set during the runtime of the path. The semantic trajectory point set is divided into fixed time windows, and local aggregation and structured segmentation processing are performed to output a trajectory segment sequence containing continuous segments, mutation segments, and drift segments.

[0064] Perform disturbance structure identification and trust change trend modeling on the trajectory segment sequence, and output the path trust attenuation factor and disturbance entropy value set;

[0065] Cross-path fusion modeling is performed by combining path trust attenuation factors and disturbance entropy value sets to construct a redundant path trust evolution structure surface. This surface is used to express the trust evolution status between paths and forms the dynamic input basis for risk modeling and risk assessment, supporting fault state identification and quantification of fault tolerance mechanism responses.

[0066] It should be further explained that the path disturbance behavior extraction is performed first. The goal of path disturbance behavior extraction is to extract time-continuous disturbance features from the path behavior trajectory, including speed, acceleration and turning trend;

[0067] Construct the perturbation velocity function, which is the first-order change in path behavior;

[0068] Among them S i (t) is the behavioral response trajectory of path i at time t. The behavioral response trajectory is the running trajectory formed by the input and output pair construction; is the perturbation velocity of path i, which represents the rate of change of the trajectory in the time domain;

[0069] Next, we construct the perturbation acceleration function, which is the second-order change of the path behavior; in is the disturbance acceleration of path i, which indicates the aggravation or attenuation trend of trajectory change;

[0070] Construct a perturbation directionality function, which represents the turning frequency of the path; Among them F i (t) is the directional expression of the path perturbation, which indicates the turning angle frequency of the trajectory at the local moment. The larger the frequency, the more drastic the change. In addition, the denominator of the formula introduces The purpose is to suppress the unstable amplification of the disturbance frequency at high speeds. The arctan(·) is used to express the steering angle variation trend of the path disturbance under the combined influence of velocity and acceleration. It measures the degree of directional fluctuation of the trajectory at a certain moment.

[0071] The disturbance intensity is modeled by integrating the nonlinear changes of speed, angular frequency, and path behavior to form a unified disturbance intensity function;

[0072] First, construct the perturbation intensity function:

[0073] where Ψ i (t) is the disturbance intensity function of path i, which is used to reflect the comprehensive degree of instantaneous change of its running trajectory;

[0074] In i (t) The first term in the formula: velocity squared, reflecting the intensity of trend fluctuations;

[0075] In i The second term in (t): the sine square of the behavior trajectory, reflecting its periodic disturbance in the amplitude domain;

[0076] In i The third term in the formula (t): the square of the angular frequency, which reflects the strength of the turning trend.

[0077] Then, the perturbation intensity is nonlinearly converted into a complexity metric to provide a compression index for subsequent trust calculations and construct a perturbation complexity function: Ω i (t) = ln(1 + Ψ i (t))tan(Ψ i (t)); where Ω i (t) is the perturbation complexity function of path i; ln(1+Ψ i (t)) represents the growth rate of the logarithmic compression disturbance value to avoid amplification of extreme disturbances; tan(Ψ i (t)) is used to improve the nonlinear perception of small disturbance changes;

[0078] Construct a disturbance entropy function, which represents the residual effect of path disturbance within a time window τ and reflects the degree of disturbance entropy accumulation. The disturbance entropy expression of the disturbance entropy function is: Among them E i (t) is the path perturbation entropy, which represents the evolution of the path perturbation within the window; exp(-t / τ) is the time weight factor, which is used to control the attenuation of the perturbation effect within the time window τ;

[0079] Construct a trust adjustment factor to measure the "stability" of the path's current state's response to the dynamics, and use it to adjust the trust value fluctuation range; Φ i (t) = cos(E i (t))+sin(Ω i (t)); where Φ i(t) is the trust adjustment factor of path i; in addition, the above formula is used to express the nonlinear response to the dynamic complexity and entropy residue, reflecting the anti-disturbance stability of the path.

[0080] Based on the disturbance entropy and the adjustment factor, the current trust value of the path at time t is expressed, and the path trust function is modeled; the path trust function is expressed as:

[0081] T i (t) = exp(-E i (t)·Φ i (t));

[0082] Where T i (t) is the path trust value; the exponential decay structure in the formula is used to reflect the trust residual of the path under the influence of disturbance entropy and anti-disturbance factor;

[0083] Build a trust structure plane. The goal of building a trust structure plane is to express the mapping of trust states between paths on the structural domain for path selection and fault tolerance comparison.

[0084] The path structure surface function is expressed as: i (t) = log(1 + [T i (t)] 2 ·sin(Ω i (t)));

[0085] where Λ i (t) is the path trust structure surface; the above formula expresses the relative state of the current path in the structure space driven by the trust value combined with the perturbation complexity.

[0086] Taking the path trust evolution structure as input, we perform pattern recognition and variation density statistics to output a set of risk factors that represent path stability.

[0087] The risk factor set is input into the directional consistency analysis process, the intersection of drift directions between paths is extracted, and the path resonance drift intersection point set is output;

[0088] The resonance drift intersection set is input into the risk expression construction process, the graph generation operation is performed, and the trust gradient expression graph is output; the trust gradient expression graph is structurally fused with the current path execution graph, and a diagnostic starting point layer with response trigger characteristics is output.

[0089] Taking the diagnosis response starting point layer as input, the path spatial location mapping and trust weight threshold state recognition process are executed, and the set of paths to be stripped containing path identification, function type and state label is output;

[0090] Taking the set of paths to be stripped as input, the operations of disconnecting the master connection, canceling the path fusion permission, and clearing the status cache are performed in sequence. The structural position, connection dependency, current status, and cancellation feedback information of each path are extracted to generate a path stripping status information table. The spatial location distribution, functional type classification, and stripping mark status recorded in the stripping status information table are used to construct a path resource distribution map in the stripping state.

[0091] Taking the path resource distribution map in the stripped state as input, the algorithm performs a consistency screening process based on semantic mapping relationships according to the functional label, semantic behavior characteristics, and historical response sequence of each path to be replaced, and outputs a set of candidate replacement paths that meet functional equivalence, semantic projection overlap, and behavioral convergence stability.

[0092] The candidate replacement path set is subjected to semantic fitting evaluation and boundary consistency verification. The semantic fitting evaluation and boundary consistency verification include path semantic vector reconstruction, behavior boundary response rate calculation and fusion compatibility verification, and the adaptation path that meets the replacement conditions is output.

[0093] Perform a virtual load response test on the adaptation path. The virtual load response test monitors its behavior convergence and response stability based on the simulated input flow required by the current master control scheduling, and outputs a verification pass path that meets the convergence conditions.

[0094] Connect the verified path to the main path execution structure, synchronize the execution path structure replacement with the result, register the path replacement time point, replacement path identifier and evolution trust value to the trust evolution record system, and output the updated path execution structure diagram;

[0095] Construct semantic feature difference function:

[0096] The semantic feature difference function is used to measure the structural difference between the semantic feature vectors mapped by path i and candidate path j at the same time t. The Euclidean distance is calculated by taking the square root of the vector difference. It reflects the absolute deviation between the two paths in the semantic space and is a quantitative expression of the degree of path structural heterogeneity.

[0097] Among them U i (t) represents the semantic feature vector of path i in the semantic space, which is derived from the projection of the path behavior trajectory; U j (t): represents the semantic feature vector of candidate path j; the square of the difference reflects the feature deviation of the path in each dimension, and the root of the square sum forms the distance between vectors; D output by the semantic feature difference function ij (t) represents the structural difference index;

[0098] Construct a boundary response difference function, which is used to measure the difference in responsiveness between paths i and j under the "semantic boundary input condition." The boundary response rate describes the sensitivity of the path output when it approaches its functional limit. The formula is formed by taking the square root of the difference between the two to form a response capability difference index.

[0099] where R i is the boundary response rate of path i; R j is the boundary response rate of path j; ij The square root of the difference in the formula reflects the degree of behavioral sensitivity deviation; R ij is the difference in response stability;

[0100] Constructing semantic direction angle function The semantic direction angle function is used to determine the consistency of the semantic evolution direction of two paths. It calculates the angle between the semantic feature vectors. If the angle is small, it means that the paths are consistent in semantic trends and are more suitable for replacement paths. It is expressed using the vector cosine formula, and the inverse cosine is used to obtain the angle.

[0101] U i (t)·U j (t) represents the vector dot product, which is used to reflect the degree of alignment of the directions of two vectors; ||U i (t)||,||U j (t)|| respectively represent the modulus length of the two vectors, which are normalized. The closer the cosine value is to 1, the more consistent the direction is. The final output is Θ ij (t) represents the path directionality deviation;

[0102] Constructing path behavior difference intensity function;

[0103] The path behavior difference intensity function is used to integrate three difference indicators: structural difference, boundary response difference, and semantic direction difference into a unified behavior difference intensity indicator. The square form is used to enhance the dominance of larger deviations over the total amount, forming a total difference intensity value as the input of the scoring function.

[0104] Among them D ij (t) 2 is the path structure difference strength item; is the boundary response difference intensity term; Θ ij (t) 2 is the semantic trend difference intensity item; the final output is Ξ ij (t), represents the combined behavioral difference index, which serves as the scoring basis for the following formula;

[0105] Construct a semantic compatibility scoring function:

[0106] The semantic compatibility scoring function converts behavioral difference indicators into scoring values. The smaller the difference, the higher the score. The structure of "inverse suppression + logarithmic compression" is used to ensure that the score increases significantly when the difference is small and decreases slowly when the difference is large, forming a nonlinear response of the scoring curve.

[0107] The denominator is 1+Ξ ij (t) represents the total difference, the larger the difference, the more dissimilar the paths are; the numerator is always 1, which is used to form an inverse suppression form; the logarithmic function in the formula is used to smooth the distribution and suppress score mutations; C ij (t) Score semantic compatibility;

[0108] Construct a path suitability scoring function as the final output;

[0109] The path suitability scoring function is C ij (t) and Ξ ij (t) are jointly mapped to the final fitness score of the alternative path; the sine function sensitively captures the high score interval, and the cosine function suppresses the difference interval, and the combination forms an alternative priority indicator with nonlinear response; sin(C ij (t)) indicates enhanced sensitive response to high-scoring segments; Indicates that the inhibition score of high-difference pathways is amplified; A ij (t) is the comprehensive adaptability score, which is used for path screening and sorting.

[0110] The trust trajectory of the stripped path before replacement, the corresponding risk factors, and the verified replacement path information are used as input. The path behavior archiving, label standardization, and evolution stage marking processes are performed to output a complete degradation learning sample set.

[0111] The trust trajectory of the current running path is aligned with the degradation learning sample set for feature vector analysis, trend direction fitting and error interval calculation are performed, and the degradation trend label of the current path is output.

[0112] Input the degradation trend label of the current path into the path trust intervention mechanism, perform trust limit adjustment, access permission reduction and alternative path scheduling initialization operations, and output the corresponding path trust control instruction set;

[0113] Taking the path trust control instruction set as input, the system executes the recording and archiving process of the path behavior intervention results. The recording and archiving process includes the control response data, the path switching process and the post-intervention status identification, and outputs the knowledge graph used to update the path evolution logical structure.

[0114] Construct path trajectory fitting error function;

[0115] The path trajectory fitting error function is used to calculate the trust trajectory T of the current path i i (t) Trust trajectory L of the historical degradation learning sample path k k (t); the error index in the form of Euclidean distance is obtained by taking the square root of the difference, which is a direct reflection of the degree of fitting of the degradation trend; where T i (t) is the path trust value; L k (t) is the trust trajectory of the historical learning sample path k; δ i (t) is the fitting error, which is used to reflect whether path i is moving closer to the historical degradation learning sample;

[0116] Construct an intervention level function, which is used to calculate the intervention level according to the fitting error δ i (t) and the set error threshold θ d Determine whether the current state of the path should trigger trust intervention and quantify its intervention level; the intervention level function uses logarithmic smoothing of the contrast factor to suppress extreme changes while enhancing the sensitivity control of the intervention level;

[0117] The error threshold θ d is the fitting error threshold of path trust intervention; B i (t) is the current intervention level of path i; if δ i (t)>θ d , then B i (t)>log2, indicating that intervention is needed;

[0118] A trust control instruction scoring function is constructed. This function is used to jointly map the intervention level and error fluctuation characteristics into a control instruction score value. The score value serves as the input basis for output control parameters (including trust limit adjustment, access permission reduction, and alternative path scheduling). The sine function is used to respond to level fluctuations, and the cosine function is used to process error intensity, forming a nonlinear but continuous control segment.

[0119] I i (t) = sin(B i (t))+τ c ·cos(δ i (t));

[0120] Among them I i (t) is the trust control score, which is used to determine the intervention intensity and implementation scope; τ c is the intervention intensity magnification coefficient, which is used to adjust the control instruction level; sin(·) and cos(·) are used to capture the level fluctuation and error trend stability respectively; Ii Higher output values ​​of (t) indicate the need for stronger control or alternative action;

[0121] The following further explanations are needed for the above:

[0122] 1. Fit and compare the current trust trajectory with the historical degradation sample trajectory;

[0123] 2. Get the trajectory fitting error δ i (t);

[0124] 3. Input the error into the intervention level function and output the current trust risk level;

[0125] 4. Jointly map the intervention level and error trend into the control score I i (t);

[0126] 5. Rating value driven: trust limit adjustment, access permission downgrade, and path replacement scheduling initialization.

[0127] An industrial Internet equipment fault diagnosis system includes a path extraction module, a redundant path module, a calculation module, a modeling semantic boundary module, a behavior graph module, a trust relationship module, and a trust structure surface module;

[0128] The path extraction module is used to input the original control instruction sequence from the industrial equipment control bus into the structured parser to perform syntax classification and function attribution extraction, and output a path set with logical uniqueness;

[0129] The redundant path module is used to input the path set into the path label engine to perform structure classification mapping, and output a redundant logical path set including the main execution path, the secondary execution path, and the mirror verification path;

[0130] The operation module is used to input the redundant logic path set and the standard working condition sequence into the execution environment, execute the fixed input operation process, and output the path output data corresponding to the path;

[0131] The modeling semantic boundary module is used to input the path output data sequence into the trend modeling mechanism to perform continuity analysis and boundary identification, and output a semantic response profile with boundary constraint characteristics;

[0132] The behavior graph module is used to input the input-output pairs consisting of the standard working condition sequence and the path output data sequence, and input them together with the semantic response profile into the semantic projection builder, perform semantic mapping modeling, and output the semantic behavior graph;

[0133] The trust relationship module is used to input the semantic behavior graph into the trust initialization model to perform consistency comparison between paths and output the initial path trust weight graph;

[0134] The trust structure plane module maps input flows to redundant paths and tracks their temporal semantic trajectories to form a structural plane that expresses the evolutionary state of trust between paths, providing dynamic input for subsequent risk modeling and risk assessment, thereby establishing a quantifiable evolutionary foundation for fault state identification and fault-tolerant mechanism response.

[0135] It should also be noted that this solution first performs a structural analysis on the original control instructions from the control bus to form a set of logically unique paths. It then uses a path label engine to divide these into a primary path, a secondary path, and a mirrored verification path, forming a set of redundant logical paths with structural redundancy.

[0136] To achieve baseline semantic modeling, redundant paths and standard operating condition sequences are input into the execution environment, where fixed input processes are executed to obtain path output data. This data sequence is then used to construct a semantic response profile through a trend modeling mechanism. This profile is then fed into a semantic projection builder, where it is combined with the original input and output pairs to generate a semantic behavior graph. The semantic behavior graph, as the only representation of the path semantics, is fed into a trust initialization model. This model then compares the consistency between paths to produce an initial trust weight graph between paths, which serves as the basis for subsequent evolution monitoring.

[0137] After entering the runtime phase, the system maps the control input stream collected in real time to the established redundant path set, performs functional operations, and obtains the output data of the path runtime phase. After the output data is structurally aligned, the semantic trajectory point set of the runtime phase is generated through the dynamic semantic projection process.

[0138] The trajectory point set is divided into time windows and aggregated to form a trajectory segment sequence containing continuous segments, sudden changes, and drift segments. The perturbation structure recognition mechanism is used to extract multi-dimensional features of the path, such as perturbation velocity, perturbation acceleration, and direction turning trend, to form perturbation intensity function and perturbation complexity function, and the perturbation entropy is constructed by superimposing the time weight factor.

[0139] Based on perturbation complexity and perturbation entropy, a path trust function is constructed. Furthermore, a trust structure surface is constructed through path group structure mapping. This serves as a quantifiable map that dynamically reflects the evolutionary trend of trust status between paths and is the core input for risk modeling and fault-tolerant control. Perturbation patterns are extracted from behavioral trajectories and a trust evolution function is constructed, enabling early identification of potential anomalies in seemingly normal paths. The path trust structure surface is input into a pattern recognition and variation density statistical mechanism to identify the stability risk factors of each path at the trust level. Directional consistency analysis is then performed to extract the directional intersection points of trust drift trends between multiple paths and determine the "resonance drift intersection point."

[0140] Based on these intersections, a trust gradient expression map is constructed and integrated with the system's current execution path map to output a "diagnostic starting point layer" with response triggering capabilities. This layer identifies the set of paths in the current system that are most likely to cause fault-tolerant structural changes and their triggering locations. This stage binds structural behavior variations to the spatial path execution map to form a risk distribution map of the path hierarchy, enabling pre-trigger location of potential fault points.

[0141] For the candidate stripping paths in the diagnosis starting layer, the spatial location mapping, function identification and trust status judgment process are executed to strip the path structure and output a resource distribution map. After judging the stripping map based on function labels, semantic behavior characteristics and response sequence consistency, candidate alternative paths are screened out.

[0142] After candidate paths undergo multi-dimensional verification, including semantic fit evaluation (semantic vector differences), behavioral boundary response rate differences, and semantic direction angles, an adapted path with behavioral consistency and structural equivalence is obtained. The adapted path is verified for behavioral stability through virtual load testing and integrated into the master control path structure. The synchronization points and trust values ​​during the replacement process are recorded in the evolution record system, enabling real-time replacement and seamless integration of the path structure. This phase establishes replacement links through multi-dimensional judgment, enabling the system to possess structural hot replacement capabilities and a semantic equivalence verification mechanism.

[0143] After the replacement of the alternative path is completed, the trust trajectory, risk factors, and alternative path behavior of the original path are archived into the sample set to construct a structurally complete degradation learning sample set; the current path trust trajectory will be trend-fitted and error-calculated with the learning sample to output a degradation trend label; the degradation label is input into the path trust intervention mechanism, and a trust control instruction set (such as limit adjustment, authority reduction, and alternative scheduling startup) is generated based on the fitting error intensity and fluctuation trend score; the control process has a nonlinear response structure, and the higher the score, the stronger the control; the intervention results will be archived in the form of control behavior data, path switching process and state identification, and ultimately form a knowledge graph to drive the system to form a closed loop of full-process learning and intervention mechanism for the path life cycle.

[0144] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for diagnosing faults in industrial Internet equipment, comprising: The original control instruction sequence from the industrial equipment control bus is input into the structured parser to perform syntax classification and function attribution extraction, and output a set of logically unique paths; Its characteristics are: Input the path set into the path label engine to perform structure classification mapping, and output a redundant logical path set including the main execution path, the secondary execution path, and the image verification path; The redundant logic path set and the standard working condition sequence are input into the execution environment, the fixed input operation process is executed, and the path output data corresponding to the path is output; The path output data sequence is input into the trend modeling mechanism to perform continuity analysis and boundary identification, and output a semantic response profile with boundary constraint characteristics; The input-output pairs consisting of the standard working condition sequence and the path output data sequence are input into the semantic projection builder together with the semantic response profile, and semantic mapping modeling is performed to output the semantic behavior map. Input the semantic behavior graph into the trust initialization model to perform consistency comparison between paths and output the initial path trust weight graph; By mapping input flows to redundant paths and tracing their temporal semantic trajectories, a structural surface is formed to express the evolutionary state of trust between paths, providing dynamic input for subsequent risk modeling and risk assessment, thereby establishing a quantifiable evolutionary foundation for fault state identification and fault-tolerant mechanism response.

2. The method for diagnosing faults of industrial Internet equipment according to claim 1, characterized in that: Map the control input stream collected during the equipment operation period to a redundant logic path set, perform real-time functional operations and collect path operation results, and output the output value sequence of the path operation period; The runtime output value sequence of the path is structurally aligned with the semantic behavior graph, and a dynamic semantic projection modeling process is executed to output a semantic trajectory point set during the runtime of the path. The semantic trajectory point set is divided into fixed time windows, and local aggregation and structured segmentation processing are performed to output a trajectory segment sequence containing continuous segments, mutation segments, and drift segments. Perform disturbance structure identification and trust change trend modeling on the trajectory segment sequence, and output the path trust attenuation factor and disturbance entropy value set; The path trust attenuation factor and the disturbance entropy value set are used to perform cross-path fusion modeling to construct a redundant path trust evolution structure surface to express the trust evolution status between paths, and form a dynamic input basis for risk modeling and risk assessment, supporting fault state identification and quantification of fault tolerance mechanism response.

3. The method for diagnosing faults of industrial Internet equipment according to claim 2, characterized in that: Taking the path trust evolution structure as input, we perform pattern recognition and variation density statistics to output a set of risk factors that represent path stability. The risk factor set is input into the directional consistency analysis process, the intersection of drift directions between paths is extracted, and the path resonance drift intersection point set is output; The resonance drift intersection set is input into the risk expression construction process, the graph generation operation is performed, and the trust gradient expression graph is output; the trust gradient expression graph is structurally fused with the current path execution graph, and a diagnostic starting point layer with response trigger characteristics is output.

4. The method for diagnosing faults of industrial Internet equipment according to claim 3, wherein: Taking the diagnosis response starting point layer as input, the path spatial location mapping and trust weight threshold state recognition process are executed, and the set of paths to be stripped containing path identification, function type and state label is output; Taking the set of paths to be stripped as input, the operations of disconnecting the master connection, canceling the path fusion permission, and clearing the status cache are performed in sequence. The structural position, connection dependency, current status, and cancellation feedback information of each path are extracted to generate a path stripping status information table. The spatial location distribution, functional type classification, and stripping mark status recorded in the stripping status information table are used to construct a path resource distribution map in the stripping state.

5. The method for diagnosing faults of industrial Internet equipment according to claim 4, characterized in that: Taking the path resource distribution map in the stripped state as input, the algorithm performs a consistency screening process based on semantic mapping relationships according to the functional label, semantic behavior characteristics, and historical response sequence of each path to be replaced, and outputs a set of candidate replacement paths that meet functional equivalence, semantic projection overlap, and behavioral convergence stability. The candidate replacement path set is subjected to semantic fitting evaluation and boundary consistency verification. The semantic fitting evaluation and boundary consistency verification include path semantic vector reconstruction, behavior boundary response rate calculation and fusion compatibility verification, and the adaptation path that meets the replacement conditions is output.

6. The method for diagnosing faults of industrial Internet equipment according to claim 5, characterized in that: Perform a virtual load response test on the adaptation path. The virtual load response test monitors its behavior convergence and response stability based on the simulated input flow required by the current master control scheduling, and outputs a verification pass path that meets the convergence conditions. Connect the verified path to the main path execution structure, synchronize the execution path structure replacement with the result, register the path replacement time point, alternative path identifier and evolution trust value to the trust evolution record system, and output the updated path execution structure diagram.

7. The method for diagnosing faults of industrial Internet equipment according to claim 6, characterized in that: The trust trajectory of the stripped path before replacement, the corresponding risk factors, and the verified replacement path information are used as input. The path behavior archiving, label standardization, and evolution stage marking processes are performed to output a complete degradation learning sample set. The trust trajectory of the current running path is aligned with the degradation learning sample set for feature vector analysis, trend direction fitting and error interval calculation are performed, and the degradation trend label of the current path is output.

8. The method for diagnosing faults of industrial Internet equipment according to claim 7, characterized in that: Input the degradation trend label of the current path into the path trust intervention mechanism, perform trust limit adjustment, access permission reduction and alternative path scheduling initialization operations, and output the corresponding path trust control instruction set; Taking the path trust control instruction set as input, the recording and archiving process of the path behavior intervention results is executed. The recording and archiving process includes control response data, path switching process and post-intervention status identification, and outputs a knowledge graph for updating the path evolution logical structure.

9. An industrial Internet equipment fault diagnosis system, comprising a path extraction module, a redundant path module, a calculation module, a semantic boundary modeling module, a behavior graph module, a trust relationship module, and a trust structure module, characterized by: The path extraction module is used to input the original control instruction sequence from the industrial equipment control bus into the structured parser to perform syntax classification and function attribution extraction, and output a path set with logical uniqueness; The redundant path module is used to input the path set into the path label engine to perform structure classification mapping, and output a redundant logical path set including the main execution path, the secondary execution path, and the mirror verification path; The operation module is used to input the redundant logic path set and the standard working condition sequence into the execution environment, execute the fixed input operation process, and output the path output data corresponding to the path; The modeling semantic boundary module is used to input the path output data sequence into the trend modeling mechanism to perform continuity analysis and boundary identification, and output a semantic response profile with boundary constraint characteristics; The behavior graph module is used to input the input-output pairs consisting of the standard working condition sequence and the path output data sequence, and input them together with the semantic response profile into the semantic projection builder, perform semantic mapping modeling, and output the semantic behavior graph; The trust relationship module is used to input the semantic behavior graph into the trust initialization model to perform consistency comparison between paths and output the initial path trust weight graph; The trust structure plane module maps input flows to redundant paths and tracks their temporal semantic trajectories to form a structural plane that expresses the evolutionary state of trust between paths, providing dynamic input for subsequent risk modeling and risk assessment, thereby establishing a quantifiable evolutionary foundation for fault state identification and fault-tolerant mechanism response.

Images