A multi-signal flow chart automatic construction technology for complex system FMECA

Abstract

The application discloses a kind of complex system FMECA-oriented multi-signal flow chart automatic construction technology.The FMECA-oriented multi-signal flow chart automatic construction technology is automatically constructed into multi-signal flow chart for FMECA file with non-standard and different expression, provides system intuitive fault propagation logic for maintenance troubleshooting personnel, improves fault diagnosis efficiency.The FMECA-oriented multi-signal flow chart automatic construction technology constrains the input of FMECA file by standardizing table header and designing ontology, and is divided into three steps: firstly, the fault event entity is analyzed by dictionary word segmentation and the TF-IDF value of each word segmentation is calculated to obtain the vector expression of the fault event entity;then, the fault event entity vectors are analyzed by the cosine of the weighted similarity of dynamic time warping, and the fault event entity pairs greater than a certain threshold are merged;finally, the corresponding nodes and their relationships of the multi-signal flow chart are obtained according to the standardized and simplified FMECA file under the constraint of ontology, and a complete multi-signal flow chart is constructed.

Description

Technical Field

[0001] This invention relates to knowledge-driven fault diagnosis methods, particularly to an automatic multi-signal flow graph construction method based on FMECA files in systems where fault diagnosis knowledge writing lacks certain standards. Background Technology

[0002] In complex, difficult-to-model, or data-scarce systems, knowledge-driven fault diagnosis methods utilize the establishment of a knowledge model of the system for fault detection, isolation, and identification. Specifically, during the design phase, knowledge-driven fault diagnosis methods oriented towards system architecture include FTA (Fault Tree Analysis), FMECA (Fault Mode, Effects, and Criticality Analysis), and multi-signal flow graphs.

[0003] (1) FTA uses logic diagrams to illustrate the relationship between failures and underlying causes. As a top-down analysis method, FTA starts from system failures or unsafe conditions (top events) and gradually decomposes them into more basic failure causes (basic events), and uses logic gates to represent the causal relationships between events. FTA can clearly identify critical failure paths and locate single or multiple basic events, but its construction process is complex and cannot exhaust all possible basic events.

[0004] (2) FMECA takes all possible failure modes of a system as its research object. Starting from individual components, it analyzes the failure information of each component and expands from the bottom up to the system level. Finally, it determines the impact of each possible failure mode on the system and summarizes it into an FMECA table. Therefore, FMECA can clarify the impact of each failure mode on the system or equipment. However, FMECA tables in complex systems are numerous and large. The global correlation between each component and each failure mode is easily obscured. Moreover, manually written FMECA data has strong differences and poor standardization, which increases the difficulty for maintenance and troubleshooting personnel to diagnose faults.

[0005] (3) Multi-signal flow graphs, as a graphical fault diagnosis model, describe the relationship between system components and test logic through directed graphs. In actual modeling, this model defines each fault mode as a "signal," and the positive and abnormal states of the "signals" are monitored by "measurement points." It utilizes the dependencies between signals and measurement points, signals and components, and components to construct the system's signal transmission model. Multi-signal flow graphs improve the efficiency of knowledge diagnosis for complex systems by focusing on the modeling of signal transmission paths involved in the fault subject and failure modes. However, because multi-signal flow graph modeling requires diagnostic and troubleshooting personnel to have a thorough understanding of the system principles, manual modeling is difficult, making its practical application challenging.

[0006] Among the aforementioned knowledge-driven fault diagnosis methods for complex system structures, FTA can clearly identify critical fault paths, but its construction process is complex and cannot cover all basic events; FMECA can analyze the impact of each fault mode on the system in detail, but it involves numerous tables, global correlations are easily overlooked, and manually written data has poor standardization; Multi-Signal Flow Graphs improve fault diagnosis efficiency through graphical methods, but manual modeling is difficult, limiting its application scope. Therefore, this invention proposes an automatic multi-signal flow graph construction technology based on FMECA, combining the comprehensiveness of FMECA analysis with the intuitiveness of multi-signal flow graphs. This reduces the difficulty of constructing multi-signal flow graphs while assisting maintenance and troubleshooting personnel in understanding the global correlations between various components and fault modes, thereby improving the comprehensiveness and accuracy of fault diagnosis for complex systems. Summary of the Invention

[0007] This invention addresses the challenges of numerous, complex, and non-standardized FMECA tables in complex systems, which exacerbate the difficulty of fault diagnosis, and the difficulty of modeling multi-signal flow graphs for engineering diagnostic scenarios. It provides an automatic multi-signal flow graph construction technique for complex system FMECAs. This method can automatically construct multi-signal flow graphs from non-standardized and inconsistent FMECA files, providing maintenance and troubleshooting personnel with a systematic and structured fault propagation logic, reducing the difficulty of using FMECA files, and improving fault diagnosis efficiency.

[0008] The aforementioned automatic multi-signal flow graph construction technology for FMECA is primarily aimed at knowledge-driven fault diagnosis scenarios for complex systems. Specifically, this technology, by standardizing headers and designing ontology constraints on the input format of FMECA files, mainly comprises three core steps: word vector generation based on TF-IDF (Term Frequency-Inverse Document Frequency), entity redundancy removal based on comprehensive similarity analysis of word vectors, and automatic construction of multi-signal flow graphs based on standardized and simplified FMECA files.

[0009] The input of the standardized header and design ontology constraint FMECA file normalizes the header of the FMECA file into seven main entities, and further constrains these seven entities to three ontology categories: "equipment components," "fault events," and "measurement points," and their relationships. The TF-IDF-based word vector generation segmentes the fault event entity, calculates the TF-IDF value of each segment, and then vectorizes each entity based on the TF-IDF value of each segment. The entity redundancy removal based on word vector comprehensive similarity analysis uses the weighted cosine similarity and dynamic time warping similarity of the TF-IDF word vectors of two entities to remove redundancy and merge entities with a similarity greater than a threshold. The automatic construction of the multi-signal flow graph based on the standardized and simplified FMECA file constructs a complete multi-signal flow graph based on the standardized and simplified FMECA file and the constraints of the ontology "equipment components," "fault events," "measurement points," and their relationships.

[0010] The features of this invention are:

[0011] (1) This invention solves the problem that the strong differences in FMECA file data exacerbate the difficulty of managing and applying fault diagnosis knowledge by designing standardized FMECA file headers and body constraints;

[0012] (2) This invention proposes a word vector generation method based on TF-IDF and a comprehensive entity similarity analysis algorithm based on cosine similarity and dynamic time warping, thereby achieving unified integration of fault event entities and solving the problem of increased ambiguity in fault diagnosis caused by differences in manual descriptions in FMECA files.

[0013] (3) Based on the standardized FMECA file, this invention clarifies the causal relationship between fault events and measurement points; the propagation relationship between fault events; the attribution relationship between fault events and equipment components, and between equipment components, and automatically constructs a multi-signal flow graph, thus solving the problem of the difficulty of manual modeling of multi-signal flow graphs. Attached Figure Description

[0014] Figure 1 Overall Steps for Automatic Construction of Multi-Signal Flow Graphs for FMECA of Complex Systems

[0015] Figure 2 ontology constraints for FMECA normalized headers

[0016] Figure 3 Generation of fault event entity word vectors based on IF-IDF

[0017] Figure 4 Redundancy removal for fault event entities based on word vector-based comprehensive similarity analysis Detailed Implementation

[0018] This invention provides an automatic multi-signal flow graph construction method for complex system FMECA files. The specific implementation of this invention is described below with reference to the accompanying drawings. The automatic multi-signal flow graph construction technology for complex system FMECA standardizes the file header and designs the input format of the FMECA by constraining the ontology. The main process includes three core steps: TF-IDF-based word vector generation, entity redundancy removal based on word vector comprehensive similarity analysis, and automatic construction of multi-signal flow graphs based on standardized and simplified FMECA files, as shown in the attached drawings. Figure 1 As shown.

[0019] 1. Header normalization and ontology constraints of FMECA documents

[0020] The FMECA document header normalization and ontology constraints include normalizing the headers of manually written, non-standard FMECA forms from different sources; and constraining each header within the ontology of "equipment components," "fault events," "measurement points," and their "attribution," "propagation," and "causation" relationships based on the normalized FMECA document headers, to support the subsequent automatic construction of multi-signal flow graphs. The ontology constraints are attached. Figure 2 As shown.

[0021] 1) The standardized FMECA document header and its corresponding description are shown in the table below:

[0022] Table 1. Standardized FMECA headers and their meanings.

[0023] FMECA Document Specification Header Explanation of table header meaning Device Code The coding describes the relationship between components and the overall equipment, facilitating the analysis of hierarchy and location. Equipment Name The device name or model corresponding to the device code. Device Functions Record the functions of products and equipment. Failure Mode The state in which a product cannot work properly is divided into potential and functional failures, covering a variety of failure modes. Failure reason Factors leading to failure include environmental, human, and physicochemical changes. Failure impact The impact of the malfunction on itself, upstream products, and the overall equipment. Failure detection Methods for identifying failure points include pre- and post-detection inspections.

[0024] 2) The specific explanation of constraining the standardized FMECA header to the three types of entities, namely “equipment components”, “fault events”, and “measurement points”, and the three types of entity relationships, namely “attribution”, “propagation”, and “causation”, is shown in the table below.

[0025] Table 2. Ontology Constraints and Their Specific Explanations

[0026]

[0027] 2. Word vector generation based on TF-IDF

[0028] The TF-IDF-based word vector generation includes dictionary-based word segmentation, TF-IDF value calculation, and word vector generation. The specific process is shown in the attached figure. Figure 3 As shown.

[0029] The dictionary-based word segmentation uses the jieba Chinese word segmentation library to segment the "fault event entity" extracted from the FMECA file and obtain the word segmentation database of the current FMECA file.

[0030] The TF-IDF-based word vector generation includes the calculation of TF and IDF values. The TF value represents the frequency of a word segment in all fault event entities in the word segmentation database. The normalization formula for the TF value of word segment i is shown below:

[0031]

[0032] The IDF value represents the reverse document frequency, indicating the rarity of a word. If a word appears in more and more fault event entities, its IDF value is close to 0, meaning it cannot effectively distinguish different fault event entities; conversely, if a word appears less frequently, its IDF value is higher, and it can be used as a keyword to distinguish different fault event entities. The specific calculation formula is shown below:

[0033]

[0034] The word vectors are generated by using the product of the TF and IDF values ​​of each word in the word segmentation database to represent the TF-IDF value of each word. Fault event entities are expressed through different word segmentation combinations, and the vector representation of each fault event entity is obtained based on the IF-IDF value of each word.

[0035] 3. Entity redundancy removal based on comprehensive word vector similarity analysis

[0036] The entity redundancy removal based on word vector comprehensive similarity analysis includes the calculation of cosine similarity of fault event entity vectors, the calculation of dynamic time warping similarity, and fault entity redundancy removal based on comprehensive similarity.

[0037] The cosine similarity of the fault event entity vectors is calculated by taking the cosine value between two fault event entity vectors to represent the similarity between the two vectors. The calculation formula is shown in the following formula:

[0038]

[0039] In the formula: S1 is the TF-IDF vector representation of fault event entity 1, S2 is the TF-IDF vector representation of fault event entity 2, and cosθ is the cosine similarity between S1 and S2.

[0040] The similarity calculation based on dynamic time warping includes calculating the distance matrix of the entity vectors of two fault events, calculating the dynamic time warping distance, and calculating the normalized similarity. The distance matrix is ​​calculated by determining the Euclidean distance between each element value in the entity vectors of the two fault events. The dynamic time warping distance is calculated by finding a path starting from the first element in the upper left corner of the distance matrix and ending at the last element in the lower right corner. The path is found by moving in three directions: right, down, and lower right, minimizing the sum of the distance matrix elements traversed by the path. The minimum sum is the dynamic time warping distance. The normalized similarity is calculated using the following formula:

[0041]

[0042] In the formula: DTW(S1,S2) is the dynamic time warping distance between fault event entity vector S1 and fault event entity vector S2, len(·) represents the dimension of the vector, and dtw_sim is the word vector similarity based on dynamic time warping.

[0043] The redundancy removal of fault entities based on comprehensive similarity is achieved by assigning a weighted value w1 to the cosine similarity (cosθ) and a weighted value w2 to the dynamic time warping distance similarity (dtw_sim), where w1 + w2 = 1, and calculating the comprehensive similarity sim = w1cosθ + w2dtw_sim. Then, based on the comprehensive similarity sim, fault event entity pairs exceeding a certain threshold are merged to complete the redundancy removal. 4. Automatic construction of multi-signal flow graphs based on standardized and simplified FMECA files.

[0044] The automatic construction of multi-signal flow graphs based on standardized and simplified FMECA files involves obtaining corresponding nodes for "equipment components," "fault events," and "measurement points" in the multi-signal flow graph under ontology constraints using standardized FMECA files. Further, a de-redundant and complete FMECA file is constructed through comprehensive similarity analysis between fault event entities. Finally, a complete multi-signal flow graph model is built based on the "attribution," "propagation," and "causation" relationships between nodes. The specific steps are as follows:

[0045] 1) Based on the equipment codes, failure detection, failure modes, failure causes and failure effects in the standardized FMECA document, under the constraints of the "equipment components" and "failure events" ontology, determine the "attribution" relationship between the equipment components that constitute the system, and the "attribution" relationship of each "failure event" in each equipment component, and summarize the failure event nodes on this basis.

[0046] 2) Based on the failure modes, causes, and effects information in the standardized FMECA document, and under the ontology constraints of "failure events," determine the "propagation" relationships between each "failure event," thereby identifying each "failure event."

[0047] The direction of transmission between nodes.

[0048] 3) Based on the failure detection, failure mode, failure cause, and failure effect in the standardized FMECA document, at the "test point"

[0049] Under the ontology constraints of "fault events", test nodes are generated based on their causal relationships, and "fault events" are identified.

[0050] The direction of transmission between the "measuring point" and the "measuring point".

[0051] 4) Merge all the above nodes to construct a complete multi-signal flow graph.

Claims

1. An automatic multi-signal flow graph construction technique for complex system FMECA, characterized in that: The aforementioned automatic multi-signal flow graph construction technology for FMECA includes input formats for specification headers and design ontology constraints FMECA files. The main steps include a word vector generation step based on TF-IDF (term frequency-inverse file frequency), which involves segmenting fault event entities in the FMECA file using a dictionary, calculating the TF-IDF value of each segment, expressing fault event entities through different combinations of segmentation, and obtaining the TF-IDF vector representation of each fault event entity by combining the TF-IDF values ​​of each segmentation. The entity redundancy removal step based on word vector comprehensive similarity analysis includes calculating the cosine similarity of fault event entity vectors, calculating similarity based on dynamic time warping, and redundancy removal of fault entities based on comprehensive similarity. The calculation of the cosine similarity of fault event entity vectors characterizes the similarity between two vectors by calculating the cosine value between them, and the calculation formula is as follows: In the formula, S1 is the TF-IDF vector representation of fault event entity 1, and S2 is the TF-IDF vector representation of fault event entity 2. Let S1 and S2 be cosine similarity; the similarity calculation based on dynamic time warping includes calculating the distance matrix of the entity vectors of the two fault events, calculating the dynamic time warping distance, and normalizing the similarity calculation; the distance matrix calculation is obtained by calculating the Euclidean distance between each element value in the entity vectors of the two fault events; the dynamic time warping distance calculation finds a path starting from the first element in the upper left corner of the distance matrix and ending at the last element in the lower right corner, stipulating that the path is found by moving in three directions: right, down, and lower right, and minimizing the sum of the elements in the distance matrix traversed by the path, then the minimum sum of the elements is the dynamic time warping distance; the normalized similarity is calculated using the formula... Calculate, where, Let S1 be the dynamic time-warped distance between fault event entity vector S1 and fault event entity vector S2. Represents the dimension of a vector. The word vector similarity is based on dynamic time warping; the redundancy removal of faulty entities based on comprehensive similarity is achieved by using cosine similarity. Weighting To dynamic time-warped distance similarity Weighting ,and Calculate the overall similarity Then, based on comprehensive similarity... Merge fault event entity pairs that exceed a certain threshold to complete redundancy removal; and automatically construct multi-signal flow graphs based on simplified FMECA files.

2. The automatic multi-signal flow graph construction technology for FMECA according to claim 1, characterized in that: The standardized header and design ontology constraints of the FMECA-oriented multi-signal flow graph automatic construction technology standardize the headers of FMECA table files from different sources (which are often manually written and non-standard) into equipment code, equipment name, equipment function, failure mode, failure cause, failure effect, and failure detection. The headers are then constrained within the ontology of "equipment components," "fault events," "measurement points," and the "attribution," "propagation," and "causation" relationships between them.

3. The automatic multi-signal flow graph construction technology for FMECA according to claim 1, characterized in that: The automatic construction step of the multi-signal flow graph based on the standardized and simplified FMECA file in the FMECA-oriented automatic construction technology obtains the corresponding nodes of the multi-signal flow graph such as "equipment components", "fault events" and "measurement points" through the standardized and deredundant simplified FMECA file under ontology constraints, and constructs a complete multi-signal flow graph model according to the "attribution", "propagation" and "causality" relationships between each node.

Images