Chassis fault intelligent diagnosis method and system

Abstract

The invention discloses a chassis fault intelligent diagnosis method and system, and relates to the technical field of fault diagnosis. The method comprises the following steps: constructing a chassis part fault knowledge graph to form multi-tuple information; obtaining target diagnosis information generated by the target fault according to the multi-element group information, and identifying a fault severity level based on the LSTM network and the target diagnosis information; and sorting the fault severity levels, performing reliability evaluation on the target diagnosis information with the higher fault severity level, and handling the target fault according to the target diagnosis information when the reliability evaluation meets a preset requirement. According to the method, on the basis of establishing the knowledge graph, a disposal scheme is given through the fault severity level and reliability evaluation, the accuracy and efficiency of fault diagnosis are improved, and human errors are reduced.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of fault diagnosis, and in particular to a chassis fault intelligent diagnosis method and system. Background Art

[0002] In the existing fault diagnosis methods for intelligent chassis components of automobiles, phenomenon-based fault diagnosis mainly relies on the personal experience of technicians and consulting maintenance manuals or knowledge bases. The experience of technicians requires a lot of practice to accumulate and is difficult to theorize and share. Maintenance manuals and knowledge bases have limitations. The manual cannot cover all possible faults and lacks comprehensiveness. For knowledge bases, the traditional construction process is to manually express and store in the form of computer language, which is time-consuming and has a long cycle. Traditional knowledge bases cannot show the association between knowledge and are prone to forming knowledge islands. In addition, the maintenance case data generated in fault diagnosis, such as fault phenomena and causes, are too colloquial and difficult to store and use effectively. Therefore, there is an urgent need for a method to realize the automated construction of knowledge bases, establish knowledge associations, accumulate and reuse diagnostic knowledge, and assist technicians in fault diagnosis. Summary of the invention

[0003] The embodiment of the present invention provides a chassis fault intelligent diagnosis method and system, which, based on the establishment of a knowledge graph, provides a treatment plan through fault severity level and reliability evaluation, thereby improving the accuracy and efficiency of fault diagnosis and reducing human errors.

[0004] A first aspect of an embodiment of the present invention provides a chassis fault intelligent diagnosis method, comprising:

[0005] Construct a knowledge graph of chassis component failures to form multi-tuple information;

[0006] Obtain target diagnostic information generated by the target fault according to the multi-tuple information, and identify the fault severity level based on the LSTM network and the target diagnostic information;

[0007] The fault severity levels are sorted, and reliability evaluation is performed on target diagnosis information with higher fault severity levels. When the reliability evaluation meets preset requirements, the target fault is handled according to the target diagnosis information.

[0008] The embodiment of the present invention constructs a knowledge graph of chassis component faults, which lays a solid theoretical foundation for subsequent fault diagnosis. Through the multi-tuple information in the knowledge graph, we can accurately determine the severity level of the target fault, thereby achieving effective classification and priority sorting of fault information. Based on the reliability evaluation of the target diagnostic information with a higher fault severity level, the target fault is processed, which can effectively solve the problem of low accuracy and poor timeliness of intelligent chassis fault processing caused by differences in staff knowledge reserves.

[0009] In addition, this method is also highly versatile and extensible. The construction of knowledge graphs and the application of intelligent diagnosis technology can flexibly adapt to the fault diagnosis needs of different vehicle models and chassis components, providing a strong guarantee for the safe operation of vehicles. At the same time, with the continuous development and improvement of technology, this method can be further expanded and optimized to adapt to more complex and diverse fault diagnosis scenarios.

[0010] As an optional implementation, in the first aspect of the embodiment of the present invention, a chassis component failure knowledge graph is constructed to form multi-tuple information, including:

[0011] Collecting fault information and preprocessing text data of the fault information;

[0012] Performing text annotation on the preprocessed fault information;

[0013] Extracting key entity information based on an entity recognition model of a bidirectional gated recurrent unit conditional random field to extract key entity information from the fault information after the text is annotated;

[0014] Construct entity relationship tuples based on the acquired key entity information and corresponding association information;

[0015] The entity relationship tuple is visualized to obtain the chassis component failure knowledge graph.

[0016] The embodiment of the present invention realizes the structured and visualized representation of fault information through text preprocessing, text annotation, entity recognition model and visualization operation, providing strong support for subsequent fault diagnosis. This method not only improves the accuracy and efficiency of fault diagnosis, but also provides a more scientific basis and guidance for the repair and maintenance of chassis parts.

[0017] As an optional implementation, in the first aspect of the embodiment of the present invention, constructing an entity relationship tuple based on the acquired key entity information and the corresponding association information includes:

[0018] The fault method based on artificial rules extracts the association relationship, which specifically includes:

[0019] By observing the characteristics of chassis component failure data itself and combining the extracted key entity information, entities are connected by defining relationship sets to form the entity relationship tuple, that is, tuple information, which is at least a triple.

[0020] The embodiment of the present invention extracts association relationships through artificial rules to construct entity relationship tuples, thereby achieving deep mining and structured representation of fault information, which not only improves the accuracy and reliability of fault diagnosis, but also provides a more scientific basis and guidance for subsequent repairs and maintenance, helps to improve the reliability and safety of chassis components, and provides a more powerful guarantee for the safe operation of vehicles.

[0021] As an optional implementation, in the first aspect of the embodiment of the present invention, obtaining target diagnostic information generated by the target fault according to the multi-tuple information, and identifying the fault severity level based on the LSTM network and the target diagnostic information include:

[0022] The target diagnostic information is processed by word embedding using the Word2vec algorithm to obtain a word embedding vector;

[0023] Inputting the word embedding vector into a pre-trained LSTM network to obtain an initial fault level feature vector;

[0024] Optimizing the initial fault level feature vector through an attention mechanism to obtain a final fault level feature vector;

[0025] The final fault level feature vector is input into the output layer, and the fault severity level is obtained by calculating the Softmax function of the output layer.

[0026] The embodiment of the present invention combines the LSTM network, the attention mechanism and the Softmax function to achieve deep mining and precise identification of fault information, which can improve the accuracy and efficiency of fault diagnosis and further improve the operating reliability and safety of the vehicle.

[0027] As an optional implementation manner, in the first aspect of the embodiment of the present invention, the fault severity levels are sorted, and reliability evaluation is performed on target diagnostic information with a higher fault severity level, including:

[0028] The reference texts corresponding to a preset number of target diagnostic information ranked high in fault severity level are compared with the fault text corresponding to the target fault to obtain the similarity between the fault text and each reference text, and the similarity is used as the reliability evaluation.

[0029] The embodiment of the present invention can plan the fault management sequence by sorting the fault severity levels and performing reliability evaluation on the target diagnostic information with higher fault severity levels. It also needs to diagnose the fault causes and promptly deal with the fault information with higher severity levels. Moreover, these benchmark texts are usually extracted from historical fault data and represent typical descriptions of known fault modes. By performing a similarity comparison with the fault text corresponding to the target fault, a fault diagnosis result with higher reliability can be obtained, which helps to reduce misjudgments and missed judgments.

[0030] As an optional implementation, in the first aspect of the embodiment of the present invention, the similarity comparison between the reference texts corresponding to the preset number of target diagnostic information with the highest fault severity ranking and the fault text corresponding to the target fault includes:

[0031] In the vector space model, each reference text and fault text is mapped to obtain the corresponding mapping feature vector:

[0032] V(d j )=(t 1j ,w 1j ;t 2j ,w 2j ;…t nj ,w nj )

[0033] V(d i )=(t 1i ,w 1i ;t 2i ,w 2i ;…t ni ,w ni )

[0034] Among them, (t kj , w kj ) is the kth feature item in the fault text, t kj is the kth feature in the fault text, 1≤k≤n, n is the total number of features in the fault text or the reference text, w kj t kj The weight in the fault text, and:

[0035]

[0036] t fkj is the number of times the kth feature in the fault text appears in the fault text, d fkj is the number of times all features appear in the fault text; N is the sum of the number of fault texts and reference texts;

[0037] (t ki , w ki) is the kth feature item in the i-th benchmark text, t ki is the kth feature in the i-th benchmark text, w ki t ki The weight in the i-th base text, and:

[0038]

[0039] Among them, t fki is the number of times the kth feature in the i-th benchmark text appears in the fault text, d fki is the number of times all features appear in the i-th benchmark text;

[0040] The mapping feature vector V(d j ) and the mapping feature vector V(d i ) to do cosine similarity comparison.

[0041] The embodiment of the present invention realizes the accurate calculation of the similarity between the fault text and the reference text through the application of vector space model and cosine similarity comparison, which provides strong technical support for the intelligent diagnosis of chassis faults, not only improves the accuracy and reliability of fault diagnosis, but also helps to optimize maintenance decisions.

[0042] As an optional implementation manner, in the first aspect of the embodiment of the present invention, when the reliability evaluation meets the preset requirements, handling the target fault according to the target diagnosis information includes:

[0043] Determine a target reference text corresponding to a target similarity, wherein the target similarity is the maximum value of similarities between the fault text and all reference texts, and the maximum value is greater than or equal to a preset threshold;

[0044] The target diagnostic information corresponding to the target reference text is used as a final diagnostic result, and the target fault is handled according to the diagnostic result.

[0045] When the reliability evaluation meets the preset requirements, the embodiment of the present invention adopts a strategy of handling the target fault according to the target diagnostic information, thereby achieving a seamless connection between diagnosis and handling, improving the efficiency and accuracy of fault diagnosis and handling, and providing maintenance personnel with more clear and specific handling guidance, thereby improving the efficiency and quality of maintenance work.

[0046] In a second aspect of an embodiment of the present invention, a chassis fault intelligent diagnosis system is provided, which includes:

[0047] A construction unit is used to construct a knowledge graph of chassis component failures to form multi-tuple information;

[0048] A level identification unit, used for obtaining target diagnosis information generated by the target fault according to the multi-tuple information, and identifying the fault severity level based on the LSTM network and the target diagnosis information;

[0049] The handling unit is used to sort the fault severity levels and perform reliability evaluation on target diagnosis information with higher fault severity levels, and when the reliability evaluation meets preset requirements, handle the target fault according to the target diagnosis information.

[0050] A third aspect of an embodiment of the present invention discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the steps of the chassis fault intelligent diagnosis method disclosed in the first aspect of the embodiment of the present invention are implemented.

[0051] A fourth aspect of an embodiment of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the steps of the chassis fault intelligent diagnosis method disclosed in the first aspect of the embodiment of the present invention.

[0052] A fifth aspect of an embodiment of the present invention discloses a computer program product. When the computer program product is run on a computer, the computer is enabled to execute the steps of the chassis fault intelligent diagnosis method disclosed in the first aspect of the embodiment of the present invention.

[0053] A sixth aspect of an embodiment of the present invention discloses an application publishing platform, which is used to publish a computer program product. When the computer program product runs on a computer, the computer executes the steps of the chassis fault intelligent diagnosis method disclosed in the first aspect of the embodiment of the present invention.

[0054] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0055] The embodiment of the present invention constructs a knowledge graph through chassis component failures, and determines the fault severity level of the target fault based on the multi-tuple of the chassis component failure knowledge graph, so as to perform reliability evaluation on the target diagnostic information with a higher fault severity level, and handle the target fault accordingly. On the basis of the knowledge graph, it solves the problems of low processing accuracy and poor timeliness of intelligent chassis faults caused by differences in staff knowledge reserves through intelligent diagnosis technology, ensures the stable operation of chassis component equipment, improves the accuracy and efficiency of fault diagnosis, and ensures the safe operation of the vehicle. It is particularly suitable for fault diagnosis of complex systems and the diagnosis of complex faults that may be caused by multiple factors, and has strong versatility and extensibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flowchart of a chassis fault intelligent diagnosis method provided by an embodiment of the present invention;

[0057] Figure 2 It is a schematic diagram of a process of constructing a chassis component failure knowledge graph provided by an embodiment of the present invention;

[0058] Figure 3 is a schematic diagram of a process for identifying a fault severity level provided by an embodiment of the present invention;

[0059] Figure 4 It is a structural schematic diagram of a chassis fault intelligent diagnosis system provided by an embodiment of the present invention;

[0060] Figure 5 It is a structural schematic diagram of a motor provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical scheme and advantages of the present invention clearer, the specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. It should be noted that, under the premise of no conflict, the various embodiments described below or the various technical features can be arbitrarily combined to form a new embodiment. Except for special instructions, the materials and equipment used in this embodiment can be purchased from the market. Examples of embodiments are shown in the accompanying drawings, in which the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and the limitations of the present application cannot be understood.

[0062] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as a limitation on the present application. In the description of the present application, "plurality" means two or more, unless otherwise precisely and specifically specified.

[0063] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "connected", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, or it can be connected through an intermediary medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0064] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0065] Based on the knowledge graph, the embodiment of the present invention uses intelligent diagnosis technology to solve the problems of low accuracy and poor timeliness in processing intelligent chassis faults caused by differences in staff knowledge reserves, ensuring the stable operation of chassis components and equipment, improving the accuracy and efficiency of fault diagnosis, and ensuring the safe operation of the vehicle. The following is a detailed description.

[0066] Embodiment 1

[0067] See also Figure 1 , Figure 1 FIG. 1 is a flow chart of a chassis fault intelligent diagnosis method disclosed in an embodiment of the present invention. Figure 1 As shown, the chassis fault intelligent diagnosis method includes:

[0068] S110, construct a knowledge graph of chassis component failures to form multi-tuple information.

[0069] The knowledge graph is a special knowledge base stored in a graph. The structure of the knowledge graph mainly includes elements such as entities, entity attributes, relationships, and relationship attributes. Entities in the knowledge graph are in the form of nodes, and relationships exist in the form of directed edges. Entities are connected through relationships to eventually form a triple structure of "entity-relationship-entity", in which attributes are used to describe some characteristics of the contained entities or relationships.

[0070] The construction of knowledge graph is mainly divided into three steps: knowledge extraction, knowledge fusion and knowledge processing. Knowledge extraction mainly extracts the required entities, attributes and relationships from structured data or unstructured data. Knowledge extraction is generally divided into two categories: entity extraction and relationship extraction.

[0071] Entity extraction mainly involves extracting words that can represent entities or attributes from the fault analysis corpus, which can complete the dimension of the entity concept set in the knowledge graph. The result of entity extraction is the node in the knowledge graph. During the extraction process, it is necessary to take into account the minimum redundancy to meet the application and visualization display.

[0072] Relation extraction is mainly used to extract the semantic relationship between entities from the chassis fault analysis corpus, describing the internal relationship between each entity. It is an essential step to provide query functions and visualization.

[0073] In a preferred embodiment of the present invention, the purpose of chassis fault diagnosis is achieved by constructing a knowledge graph of chassis component faults (chassis knowledge graph for short), and the chassis knowledge graph defines relationships through entities and their attributes. Knowledge fusion mainly performs entity disambiguation and coreference resolution on the information extracted from knowledge, integrates the extracted information into a multi-tuple form, and forms a preliminary knowledge graph. Knowledge processing mainly evaluates and updates the preliminary knowledge graph to meet application requirements.

[0074] The knowledge graph is divided into a data layer and a model layer. The data layer is composed of a series of factual data, containing a large amount of factual information, namely (entity, relationship, entity) or (entity, attribute, attribute value). The model layer is used to construct entities, attributes, and relationships from data. It is the core of the knowledge graph and stores refined knowledge. Ontology libraries are widely used to construct the data layer of the knowledge graph. The knowledge graph constructed by the ontology library has less redundancy and a strong structural hierarchy.

[0075] The construction of chassis parts fault knowledge graph is mainly divided into three parts: entity extraction, relationship extraction, and knowledge storage. Figure 2 As shown, it can specifically include the following steps:

[0076] Collect fault information and preprocess the text data of the fault information. Fault information is generally fault description information, which is a kind of text data. Preprocessing mainly involves data cleaning of text data to check data consistency and process invalid values ​​and missing values.

[0077] Perform text annotation on the pre-processed fault information. The text annotation is performed manually, and text annotation can be performed on part of the fault information, mainly annotating key entity information in the text data, and the key entity information may include phenomenon entity, pointing entity and component entity.

[0078] The component entity corresponds to the specific component of the chassis; the phenomenon entity corresponds to the description of specific fault phenomena such as breakage, wear, sound, alarm number, etc.; the pointing entity is the pointing keyword corresponding to the phenomenon entity, and the keyword can point to the corresponding chassis fault phenomenon, which is obtained by segmenting the phenomenon entity and removing stop words. For example: for the phenomenon entity "wire control brake failure", the corresponding pointing entities can be: "wire control", "brake", and "failure".

[0079] Text annotation is mainly used to train the entity recognition model, so that other fault information can be used to extract the corresponding key entity information through the entity recognition model. The entity recognition model can be implemented using a bidirectional gated recurrent unit conditional random field (BiGRU-CRF).

[0080] An entity relationship tuple is constructed based on the acquired key entity information and the corresponding association information. The entity relationship tuple is constructed by extracting the association relationship based on the fault method of artificial rules, and then connecting the association relationship with the key entity information. Specifically, by observing the characteristics of the chassis component fault data itself and combining the extracted key entity information, the entities can be connected by defining a relationship set to form the entity relationship tuple, that is, to construct the tuple information of the chassis component fault. It should be noted that the tuple is at least a triple, and in order to adapt to subsequent operations, the tuple needs to include a solution to the cause of the fault, and the solution includes but is not limited to a maintenance plan, and spare parts information, etc. For example, the association relationship definition is respectively the fault location point, the cause of the fault information, and the solution corresponding to the fault cause, forming a triple information. It is also possible to define tuple information such as fault phenomenon, fault location, fault cause, maintenance plan, spare parts information, etc.

[0081] It can be understood that a fault can correspond to one or more multi-tuple groups. For example, the failure of the wire control brake can correspond to fault locations including the electronic pedal, the main control node, the brake node, etc., and each fault location can correspond to multiple causes of the fault information. For example, when the above-mentioned wire control brake failure corresponds to the fault location of the electronic pedal, the causes of the fault information can be divided into damage to the pedal mechanical structure, failure of the pedal position sensor circuit, and disconnection between the pedal and the ECU, etc.

[0082] The following is a partial list of common chassis faults, their corresponding fault locations and fault causes, to explain that each fault may correspond to many tuples, especially in complex faults (which may have multiple fault locations and / or multiple fault causes).

[0083] Fault 1: Brake-by-wire failure.

[0084] Braking failure is mainly manifested in that the car does not slow down when the driver issues a braking command while the car is driving, and there is no obvious response even after several consecutive brakings. The fault locations and causes of this type of fault are as follows:

[0085] 1.1. Fault location: electronic pedal.

[0086] 1.11. Cause of failure: The pedal mechanical structure is damaged; the pedal position sensor circuit is faulty or disconnected from the ECU, resulting in no output signal.

[0087] 1.2. Fault location: main control node.

[0088] 1.21. Cause of failure:

[0089] Hardware failure: 1) Internal electronic components are damaged due to overheating, ECU printed circuit board is broken, or mechanical impact causes the cable fixing sleeve to fall off or deform, resulting in no braking force output; 2) CPU crystal oscillator circuit, power supply circuit and other peripheral circuits are open or short-circuited, resulting in no braking force signal output; 3) Electromagnetic interference (EMC) is serious, causing signal confusion and invalid signal output, resulting in failure to brake.

[0090] Software failure: 1) Memory overflow, stack overflow, program error; 2) Operating system locked, program error; 3) Output port address is changed, instructions cannot be output from the correct port.

[0091] Communication failure: The communication connectors between nodes are in poor contact, disconnected, or their external shielding is damaged, resulting in braking failure.

[0092] Power failure: The power circuit is suddenly disconnected and there is no voltage output, causing the system to not work properly; the power output voltage is too low, making the actuator unable to brake and unable to meet the braking requirements.

[0093] 1.3. Fault location: braking node.

[0094] 1.31、Cause of failure:

[0095] Brake controller failure: control signal output. Braking cannot be achieved.

[0096] Brake actuator failure: 1) The motor cannot work normally due to open circuit / short circuit of capacitors and resistors, damage to the power tube, etc.; 2) The gears are severely worn and the power cannot be transmitted; 3) The screw fails and the power cannot be transmitted.

[0097] Fault 2: Braking by wire does not work properly.

[0098] Braking failure is mainly manifested as the driver quickly steps on the brake pedal to the bottom while the car is driving, but the car cannot slow down or stop immediately, and the braking of each wheel is obviously delayed, that is, the braking distance and braking time are too long. The fault location and causes of this type of fault are as follows:

[0099] 2.1. Fault location: electronic pedal.

[0100] 2.11. Cause of failure: The free travel of the brake pedal is too large; the pedal position / force sensor collects signals with delay. The input signal is delayed. Braking cannot be performed immediately.

[0101] 2.2. Fault location: main control node.

[0102] 2.21. Cause of failure:

[0103] Main control unit failure: The main control unit braking force control output signal is delayed and braking cannot be performed immediately.

[0104] Communication failure: The communication connectors between nodes have poor contact and the external shielding is damaged, resulting in communication delays.

[0105] Power failure: The power supply voltage is too low or unstable.

[0106] 2.3. Fault location: braking node.

[0107] 2.31、Cause of failure:

[0108] Brake controller failure: The control signal output is delayed and the actuator cannot be controlled in time for braking.

[0109] Brake actuator failure: 1) The brake clearance is too large and the clearance self-adjusting device fails; 2) The friction plate is aged or has oil stains, resulting in poor contact between the friction plate and the brake disc; 3) The friction plate is overheated and the braking performance declines, resulting in poor braking; 4) The friction plate is wet and the braking performance declines.

[0110] Fault 3: Brake-by-wire deviation:

[0111] Braking deviation is mainly manifested in that the two coaxial wheels cannot brake at the same time during braking, and the car cannot stop in the straight direction but deviates to one side. The fault location and cause of this type of fault are as follows:

[0112] 3.1. Fault location: master control node

[0113] 3.11、Cause of failure:

[0114] Main control node failure: Incorrect calculations within the main control node lead to unreasonable distribution of braking force on both sides.

[0115] Communication failure: The node interface line on one side is disconnected or short-circuited.

[0116] Power supply failure: Inconsistent voltage output, resulting in different motor operation; some motors are not powered (open circuit).

[0117] 3.2. Fault location: braking node.

[0118] 3.21. Cause of failure:

[0119] Brake controller failure: 1) There is output but it is inconsistent with the main control distribution value; 2) There is no output on one side.

[0120] Brake actuator failure: 1) The internal circuit of the motor is open or short-circuited; 2) The gears are severely worn; 3) The screw fails; 4) The friction plate is aged or sticky with oil, resulting in poor contact between the friction plate and the brake disc; 5) The friction plate is overheated and the braking performance is reduced; 6) The brake clearances of each wheel are not equal, and the degree of wear of the left and right friction plates is inconsistent; 7) The friction plate is stained with water and the braking performance is reduced.

[0121] Fault 4: Brake-by-wire spontaneous braking.

[0122] Spontaneous braking mainly occurs when the driver does not issue a braking command and the car brakes suddenly. The fault locations and causes of this type of fault are as follows:

[0123] 4.1. Fault location: electronic pedal.

[0124] 4.11. Cause of failure: The position / force sensor collects interference signals from the internal or external environment of the system. During filtering, it is unable to eliminate abnormal data and mistakenly identifies the driver's braking command, which is then transmitted to the main control unit to calculate the braking force for braking.

[0125] 4.2. Fault location: main control node.

[0126] 4.21. Cause of failure:

[0127] Program runaway failure: The program is not executed according to the normal process, but directly jumps to the braking command area for execution, resulting in a braking command being issued to execute braking when the braking conditions are not met.

[0128] Decoder failure: Translating other instructions into braking commands, resulting in spontaneous generation of erroneous braking commands.

[0129] Communication failure: The communication channel is disturbed, causing other signals to become braking signals, achieving spontaneous braking.

[0130] 4.3. Fault location: braking node.

[0131] 4.31、Cause of failure:

[0132] Brake controller failure: It receives external interference signals and mistakenly thinks it is the braking force signal sent by the main controller, and spontaneously controls the actuator to brake.

[0133] Fault 5: Brake-by-wire brake drag.

[0134] The main manifestation of brake-by-wire brake drag is that after pressing the brake pedal while driving, the brake cannot be released quickly and the brake effect is still there. The fault location and cause of this kind of fault are as follows:

[0135] 5.1. Fault location: Electronic pedal

[0136] 5.11. Cause of failure: The return spring falls off, breaks or is too soft; the position / force sensor fails, the sensor signal is delayed, and the brake cannot be released in time.

[0137] 5.2. Fault location: main control node.

[0138] 5.21. Cause of failure:

[0139] Hardware failure: Mistaking other interference signals as pedal sensor signals, keeping the brakes on and not releasing them in time

[0140] Software failure: Program runaway: Due to some reasons, the program is not executed according to the normal process, but directly jumps to the braking command area for execution, resulting in the issuance of a braking command and execution of the braking function when the braking conditions are not met.

[0141] Communication failure: interfered by other signals, maintaining the brake signal.

[0142] 5.3. Fault location: braking node.

[0143] 5.31、Cause of failure:

[0144] Brake controller failure: Hardware failure: interfered by other signals and the brakes were not released in time; Software failure: the program ran wild.

[0145] Brake actuator failure: The screw is worn, which reduces the reverse efficiency of the ball screw pair and prevents the brake shoe from returning to its original position smoothly.

[0146] The entity relationship tuple is visualized to obtain the chassis component fault knowledge graph. Ne04j graph database technology can be used to realize the storage of tuple knowledge and the visualization of graph, thereby completing the construction of chassis component fault knowledge graph. Ne04j is a conventional graph-based database that uses a graph structure to store data, making data processing more intuitive and easy to understand.

[0147] S120, obtaining target diagnosis information generated by the target fault according to the multi-tuple information, and identifying the fault severity level based on the LSTM network and the target diagnosis information.

[0148] The diagnostic information generated by the target fault is represented by word embedding vectors using the Word2vec algorithm and used as the input of the LSTM (Long ShortTerm Memory) layer for bidirectional encoding training. Then, attention is used to enhance the relevant features in fault diagnosis, and finally the Softmax function is used to recognize the fault severity level.

[0149] Fault diagnosis information is implemented in the form of tuples, so each target fault can correspond to multiple tuples. The operation of step S120 is to identify the fault severity level of all tuples corresponding to the target fault, so as to obtain the fault severity level corresponding to each tuple. Here, a triple (fault cause, fault location and treatment measure) is taken as an example for explanation and description.

[0150] Please refer to Figure 3 As shown, the fault severity level is identified based on the LSTM network and the target diagnostic information, which may include the following steps:

[0151] The target diagnostic information is word-embedded using the Word2vec algorithm to obtain a word embedding vector.

[0152] Word2vec is used as the word embedding method, and the acquired fault diagnosis information is formally represented as S f ={x1, x2, ..., x n}, where xi is the i-th word in the text Sf. For each word xi, in the training process, the fault diagnosis information needs to be preprocessed before entering the LSTM layer (here a bidirectional LSTM network is used, which can also be called a feature extraction layer). The process mainly uses the word vector matrix to represent the fault diagnosis information with the same vector. Then, the sentence S of the fault diagnosis information f , converted into a real number matrix, and the fault level information is added into it, and the obtained real number matrix is ​​passed to the next layer model.

[0153] The word embedding vector is input into a pre-trained LSTM network to obtain an initial fault level feature vector.

[0154] Each LSTM (Long Short-Term Memory) unit contains a cell state and three gate units: forget gate, input gate, and output gate. In the forward network, the input feature vector x is t , and the state h at the previous moment t-1 Calculate together to calculate the current state h t , the formula is:

[0155] i t =δ(W wi x t +W hi h t-1 +W ci c t-1 +b i )

[0156] f t =δ(W wf x t+W hf h t-1 +W cf c t-1 +b f )

[0157] Z t =tanh(W wc x t +W hc h t-1 +b c )

[0158] c t =f t c t-1 +i t z

[0159] o t =δ(W wo x t +W ho h t-1 +W co c t-1 +b o )

[0160] h t =o t tanh(c t )

[0161] Where: t is the time step; h t is the hidden state; h t-1 is the hidden state at the previous moment; i t is the input gate state; f t is the forget gate state; o t is the output gate state; c t is the cell state, c t-1 is the cell state at the previous moment; Z t is the candidate cell state; W is the corresponding weight matrix, for example W wi is the feature vector x in the input gate t The corresponding weight matrix, W wi is the hidden state h of the previous moment in the input gate t-1 The corresponding weight matrix, b i , b f , b c , b o are the bias items of the input gate state, forget gate state, candidate cell state and output gate state respectively; δ and tanh are the Sigmoid activation function and the hyperbolic tangent activation function respectively.

[0162] Since a bidirectional long-term short-term network is used, its forward long-term short-term network mainly processes the fault diagnosis information of the previous text to input the word segmentation vector x1~x t , combined with x t The previous information is used to t Perform semantic encoding and output information as h t Similarly, the reverse long-term short-term network processes the fault diagnosis information in the following text to transform the input word segmentation feature vector x n ~x t , combined with x t The following information is used to t The semantic encoding output information is h t (To distinguish, here it is rewritten as h t ′). Finally, use the obtained h t and h t ′, to represent the output information after the tth word segmentation encoding The formula is:

[0163]

[0164] Where: is the vector connection; d e is the unidirectional long-term and short-term network dimension; H is the set of output vectors, recorded as the initial fault level feature vector, which serves as the input of the next layer (attention mechanism); This means that the resulting vector belongs to the real number space, and its dimension is 2d. e ; Indicates that the resulting vector belongs to the real number space and has a dimension of 2d e n.

[0165] The initial fault level feature vector is optimized through an attention mechanism to obtain a final fault level feature vector.

[0166] In the Attention mechanism, the fault level feature vector in the fault diagnosis information text output by the bidirectional long-term short-term network at time n has an attention probability distribution of the final state. n The calculation process is as follows:

[0167]

[0168] Where: N is the number of input fault diagnosis information text sequences; is the complete hidden state sequence of fault diagnosis information at the nth moment; is the jth hidden state vector in the fault diagnosis information at the nth moment; a n is the hidden state at the nth moment for distribution of attention.

[0169] The calculation process of the final features of the attention-based fault diagnosis information text at time n is shown in the following formula:

[0170]

[0171] Where: F is the final fault level feature vector of the information text at time n. n , The higher the value of , the higher the representation power of the corresponding verb embedding is, and the more faults and important discriminative information can be obtained from the attention mechanism.

[0172] The final fault level feature vector is input into the output layer, and the fault severity level is obtained by calculating the Softmax function of the output layer.

[0173] The final fault level feature vector is input into the output layer, and the fault level type label dimension is obtained by calculating the Softmax function of the output layer, and the probability distribution thereof is shown in the following formula:

[0174]

[0175] F′=VF

[0176] Where: T is the number of fault level labels, which is set to T = 4 here, including Minor, Warning, Major, and Critical; V is the weight generated by the fault diagnosis information during the training process.

[0177] Each piece of fault diagnosis information input into the bidirectional long-term and short-term network has a corresponding label in the training set data, which is the fault severity level corresponding to the fault diagnosis information. The probability of each label is calculated by the above formula The label with the highest probability is the fault severity level caused by the fault diagnosis information.

[0178] S130, sorting the fault severity levels, and performing reliability evaluation on target diagnosis information with higher fault severity levels, and when the reliability evaluation meets preset requirements, handling the target fault according to the target diagnosis information.

[0179] Since a target fault may correspond to multiple tuples, the tuples (i.e., target diagnostic information) can be sorted according to the fault severity level, and only the target diagnostic information with a higher fault severity level can be evaluated for reliability, thereby greatly reducing the workload and improving the timeliness of fault handling.

[0180] For example, reliability evaluation can be performed only on the parts of the fault severity level that are Major and Critical. Or reliability evaluation can be performed on a preset number of target diagnostic information that are ranked high in fault severity. Here, the target diagnostic information for reliability evaluation is recorded as baseline diagnostic information, and its corresponding fault text is recorded as baseline text. The baseline text is usually extracted from historical fault data and represents a typical description of a known fault mode. It should be noted that the method of obtaining the baseline text and the fault text needs to be consistent. For example, both can be fault text data obtained by the chassis controller.

[0181] Reliability evaluation can be achieved by performing a similarity comparison between the reference text and the fault text corresponding to the target fault, obtaining the similarity between the fault text and each reference text, and using the similarity as reliability evaluation.

[0182] Specifically, it may include the following steps:

[0183] In the vector space model, each reference text and fault text is mapped to obtain the corresponding mapping feature vector:

[0184] V(d j )=(t 1j ,w 1j ;t 2j ,w 2j ;…t nj ,w nj )

[0185] V(d i )=(t 1i ,w 1i ;t 2i ,w 2i ;…t ni ,w ni )

[0186] Among them, (t kj , w kj ) is the kth feature item in the fault text, t kj is the kth feature in the fault text, 1≤k≤n, n is the total number of features in the fault text or the reference text, w kj t kj The weight in the fault text, and:

[0187]

[0188] t fkj is the number of times the kth feature in the fault text appears in the fault text, d fkj is the number of times all features appear in the fault text; N is the sum of the number of fault texts and reference texts;

[0189] (t ki , w ki ) is the kth feature item in the i-th benchmark text, t ki is the kth feature in the i-th benchmark text, w ki t ki The weight in the i-th base text, and:

[0190]

[0191] Among them, t fki is the number of times the kth feature in the i-th benchmark text appears in the fault text, d fki is the number of times all features appear in the i-th benchmark text;

[0192] The mapping feature vector V(d j ) and the mapping feature vector V(d i ) is used for similarity comparison, such as cosine similarity comparison. When the maximum value of the similarity between the fault text and all the benchmark texts is greater than or equal to a preset threshold, the target diagnostic information corresponding to the benchmark text corresponding to the maximum value (recorded as the target benchmark text) is used as the final diagnostic result, and the target fault is handled according to the handling measures of the diagnostic result.

[0193] In other embodiments, if there are multiple equal maximum values, and these maximum values ​​are greater than or equal to the preset threshold, the target faults are handled respectively according to the diagnosis results corresponding to these maximum values. If the maximum value is less than the preset threshold, it means that the fault severity level of the target fault is low and no processing is required for the time being.

[0194] Embodiment 2

[0195] See also Figure 4 , Figure 4 Schematic diagram of the structure of a chassis fault intelligent diagnosis system disclosed in an embodiment of the present invention. Figure 4 As shown, the chassis fault intelligent diagnosis system may include:

[0196] A construction unit 210 is used to construct a knowledge graph of chassis component failures to form multi-tuple information;

[0197] A level identification unit 220, configured to obtain target diagnosis information generated by a target fault according to the multi-tuple information, and identify a fault severity level based on an LSTM network and the target diagnosis information;

[0198] The handling unit 230 is used to sort the fault severity levels, and perform reliability evaluation on target diagnosis information with higher fault severity levels, and handle the target fault according to the target diagnosis information when the reliability evaluation meets preset requirements.

[0199] Preferably, the construction unit 210 may include:

[0200] Collecting fault information and preprocessing text data of the fault information;

[0201] Performing text annotation on the preprocessed fault information;

[0202] Extracting key entity information based on an entity recognition model of a bidirectional gated recurrent unit conditional random field to extract key entity information from the fault information after the text is annotated;

[0203] Construct entity relationship tuples based on the acquired key entity information and corresponding association information;

[0204] The entity relationship tuple is visualized to obtain the chassis component failure knowledge graph.

[0205] Constructing entity relationship tuples based on the acquired key entity information and corresponding association information may include:

[0206] The fault method based on artificial rules extracts the association relationship, which specifically includes:

[0207] By observing the characteristics of chassis component failure data itself and combining the extracted key entity information, entities are connected by defining relationship sets to form the entity relationship tuple, that is, tuple information, which is at least a triple.

[0208] Preferably, the level identification unit 220 may include:

[0209] The target diagnostic information is processed by word embedding using the Word2vec algorithm to obtain a word embedding vector;

[0210] Inputting the word embedding vector into a pre-trained LSTM network to obtain an initial fault level feature vector;

[0211] Optimizing the initial fault level feature vector through an attention mechanism to obtain a final fault level feature vector;

[0212] The final fault level feature vector is input into the output layer, and the fault severity level is obtained by calculating the Softmax function of the output layer.

[0213] Preferably, the processing unit 230 may include:

[0214] The reference texts corresponding to a preset number of target diagnostic information ranked high in fault severity level are compared with the fault text corresponding to the target fault to obtain the similarity between the fault text and each reference text, and the similarity is used as the reliability evaluation.

[0215] Comparing the similarity between the reference texts corresponding to the preset number of target diagnostic information items ranked higher in fault severity and the fault texts corresponding to the target faults may include:

[0216] In the vector space model, each reference text and fault text is mapped to obtain the corresponding mapping feature vector:

[0217] V(d j )=(t 1j ,w 1j ;t 2j ,w 2j ;…t nj ,w nj )

[0218] V(d i )=(t 1i ,w 1i ;t 2i ,w 2i ;…t ni ,w ni )

[0219] Among them, (t kj , w kj ) is the kth feature item in the fault text, t kj is the kth feature in the fault text, 1≤k≤n, n is the total number of features in the fault text or the reference text, w kj t kj The weight in the fault text, and:

[0220]

[0221] t fkj is the number of times the kth feature in the fault text appears in the fault text, d fkj is the number of times all features appear in the fault text; N is the sum of the number of fault texts and reference texts;

[0222] (t ki , w ki ) is the kth feature item in the i-th benchmark text, t ki is the kth feature in the i-th benchmark text, w ki t ki The weight in the i-th base text, and:

[0223]

[0224] Among them, t fki is the number of times the kth feature in the i-th benchmark text appears in the fault text, d fki is the number of times all features appear in the i-th benchmark text;

[0225] The mapping feature vector V(d j ) and the mapping feature vector V(d i ) to do cosine similarity comparison.

[0226] When the reliability evaluation meets the preset requirements, handling the target fault according to the target diagnosis information may include:

[0227] Determine a target reference text corresponding to a target similarity, wherein the target similarity is the maximum value of similarities between the fault text and all reference texts, and the maximum value is greater than or equal to a preset threshold;

[0228] The target diagnostic information corresponding to the target reference text is used as a final diagnostic result, and the target fault is handled according to the diagnostic result.

[0229] Embodiment 3

[0230] See also Figure 5 , Figure 5 FIG. 2 shows a schematic diagram of the structure of an electronic device that can be used to implement an embodiment of the present invention. Figure 5 As shown, the electronic device includes at least one processor 310, and a memory connected to the at least one processor 310, such as a ROM (read-only memory) 320, a RAM (random access memory) 330, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 310 can perform various appropriate actions and processes according to the computer program stored in the ROM 320 or the computer program loaded from the storage unit 380 to the random access memory RAM 330. In the RAM 330, various programs and data required for the operation of the motor can also be stored. The processor 310, the ROM 320, and the RAM 330 are connected to each other via a bus 340. An I / O (input / output) interface 350 is also connected to the bus 340.

[0231] A number of components in the motor are connected to the I / O interface 350, including: an input unit 360, such as a keyboard, a mouse, etc.; an output unit 370, such as various types of displays, speakers, etc.; a storage unit 380, such as a disk, an optical disk, etc.; and a communication unit 390, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 390 allows the motor to exchange information / data with other devices through a computer network such as the Internet or / and various telecommunication networks.

[0232] The processor 310 may be any general or / and dedicated processing component with processing and computing capabilities. Some examples of the processor 310 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 310 executes one or more steps of a chassis fault intelligent diagnosis method described in the first embodiment above.

[0233] In some embodiments, a chassis fault intelligent diagnosis method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 380. In some embodiments, part or all of the computer program may be loaded or / and installed on the motor via the ROM 320 and / or the communication unit 390. When the computer program is loaded into the RAM 330 and executed by the processor 310, one or more steps of the chassis fault intelligent diagnosis method described in the first embodiment above may be performed. Alternatively, in other embodiments, the processor 310 may be configured to perform a chassis fault intelligent diagnosis method in any other appropriate manner (e.g., by means of firmware).

[0234] The above is a detailed introduction to a chassis fault intelligent diagnosis method and system disclosed in the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A chassis fault intelligent diagnosis method, characterized in that: It includes the following steps: Construct a knowledge graph of chassis component failures to form multi-tuple information; Obtain target diagnostic information generated by the target fault according to the multi-tuple information, and identify the fault severity level based on the LSTM network and the target diagnostic information; The fault severity levels are sorted, and reliability evaluation is performed on target diagnosis information with higher fault severity levels. When the reliability evaluation meets preset requirements, the target fault is handled according to the target diagnosis information.

2. The chassis fault intelligent diagnosis method according to claim 1, characterized in that: Construct a knowledge graph of chassis component failures to form multi-tuple information, including: Collecting fault information and preprocessing text data of the fault information; Performing text annotation on the preprocessed fault information; Extracting key entity information based on an entity recognition model of a bidirectional gated recurrent unit conditional random field to extract key entity information from the fault information after the text is annotated; Construct entity relationship tuples based on the acquired key entity information and corresponding association information; The entity relationship tuple is visualized to obtain the chassis component failure knowledge graph.

3. The chassis fault intelligent diagnosis method according to claim 2, characterized in that: Construct entity relationship tuples based on the acquired key entity information and corresponding association information, including: The fault method based on artificial rules extracts the association relationship, which specifically includes: By observing the characteristics of chassis component failure data itself and combining the extracted key entity information, entities are connected by defining relationship sets to form the entity relationship tuple, that is, tuple information, which is at least a triple.

4. The chassis fault intelligent diagnosis method according to claim 1, characterized in that: Obtaining target diagnostic information generated by the target fault according to the multi-tuple information, and identifying the fault severity level based on the LSTM network and the target diagnostic information, including: The target diagnostic information is processed by word embedding using the Word2vec algorithm to obtain a word embedding vector; Inputting the word embedding vector into a pre-trained LSTM network to obtain an initial fault level feature vector; Optimizing the initial fault level feature vector through an attention mechanism to obtain a final fault level feature vector; The final fault level feature vector is input into the output layer, and the fault severity level is obtained by calculating the Softmax function of the output layer.

5. The chassis fault intelligent diagnosis method according to claim 1, characterized in that: The fault severity levels are sorted, and reliability evaluation is performed on target diagnostic information with a higher fault severity level, including: The reference texts corresponding to a preset number of target diagnostic information ranked high in fault severity level are compared with the fault text corresponding to the target fault to obtain the similarity between the fault text and each reference text, and the similarity is used as the reliability evaluation.

6. The chassis fault intelligent diagnosis method according to claim 5, characterized in that: The similarity comparison is performed between the reference texts corresponding to the preset number of target diagnostic information with the highest fault severity ranking and the fault text corresponding to the target fault, including: In the vector space model, each reference text and fault text is mapped to obtain the corresponding mapping feature vector: V(d j )=(t 1j ,w 1j ;t 2j ,w 2j ;…t nj ,w nj ) V(d i )=(t 1i ,w 1i ;t 2i ,w 2i ;…t ni ,w ni ) Among them, (t kj , w kj ) is the kth feature item in the fault text, t kj is the kth feature in the fault text, 1≤k≤n, n is the total number of features in the fault text or the reference text, w kj t kj The weight in the fault text, and: t fkj is the number of times the kth feature in the fault text appears in the fault text, d fkj is the number of times all features appear in the fault text; N is the sum of the number of fault texts and reference texts; (t ki , w ki ) is the kth feature item in the i-th benchmark text, t ki is the kth feature in the i-th benchmark text, w ki t ki The weight in the i-th base text, and: Among them, t fki is the number of times the kth feature in the i-th benchmark text appears in the fault text, d fki is the number of times all features appear in the i-th benchmark text; The mapping feature vector V(d j ) and the mapping feature vector V(d i ) to do cosine similarity comparison.

7. The intelligent diagnosis method for chassis fault according to claim 5, characterized in that: When the reliability evaluation meets the preset requirements, the target fault is handled according to the target diagnosis information, including: Determine a target reference text corresponding to a target similarity, wherein the target similarity is the maximum value of similarities between the fault text and all reference texts, and the maximum value is greater than or equal to a preset threshold; The target diagnostic information corresponding to the target reference text is used as a final diagnostic result, and the target fault is handled according to the diagnostic result.

8. A chassis fault intelligent diagnosis system, characterized in that: It includes: A construction unit is used to construct a knowledge graph of chassis component failures to form multi-tuple information; A level identification unit, used for obtaining target diagnosis information generated by the target fault according to the multi-tuple information, and identifying the fault severity level based on the LSTM network and the target diagnosis information; The handling unit is used to sort the fault severity levels and perform reliability evaluation on target diagnosis information with higher fault severity levels, and when the reliability evaluation meets preset requirements, handle the target fault according to the target diagnosis information.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the intelligent diagnosis method for chassis faults as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: It stores a computer program, wherein the computer program enables a computer to execute the chassis fault intelligent diagnosis method according to any one of claims 1 to 7.

Images