CN117742304A - Fault diagnosis method and system for crankshaft double-top vehicle control system - Google Patents

Abstract

The embodiment of the application relates to the technical field of artificial intelligence, in particular to a fault diagnosis method and system of a crankshaft double-top-vehicle control system. Meanwhile, the method also utilizes a trend expression element mining model to mine element vectors of a plurality of fault trend labels, and generates trend expression element vectors. By comparing the matching coefficients of the two types of vectors, the method can accurately identify the current fault trend of the system.

Description

Fault diagnosis method and system for crankshaft double-top vehicle control system

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a fault diagnosis method and system of a crankshaft double-top vehicle control system.

Background

In the field of fault diagnosis of a crankshaft double-top vehicle control system, the traditional fault diagnosis method often depends on manual experience and simple threshold judgment, and various potential faults of the system are difficult to accurately and comprehensively diagnose. With the development of sensor technology, data processing technology and machine learning algorithms, data-driven fault diagnosis methods are becoming research hotspots.

However, existing data-driven fault diagnosis methods often only focus on a single type of data (e.g., sensor detection data) when processing historical operating data of a crankshaft dual-top vehicle control system, and ignore important information such as control output data and actuator state data. This results in the fact that these methods may not accurately capture all of the characteristics of the fault when diagnosing certain complex faults, thereby reducing the accuracy of the diagnosis.

Furthermore, existing methods typically consider only fixed labels corresponding to known fault types when determining fault trends, and lack a flexible, scalable way to handle new fault types that are continuously emerging. This limits the applicability and versatility of these methods in practical applications.

Disclosure of Invention

In order to improve the technical problems in the related art, the application provides a fault diagnosis method and system of a crankshaft double-top vehicle control system.

In a first aspect, an embodiment of the present application provides a fault diagnosis method for a crankshaft dual-top vehicle control system, which is applied to an artificial intelligent fault diagnosis system, and the method includes: acquiring historical operation data to be diagnosed and X fault trend labels, wherein X is an integer not smaller than 1, and the historical operation data to be diagnosed comprises sensor detection data, control output data and actuator state data of a crankshaft double-top-vehicle control system; taking the historical operation data to be diagnosed as a processing object of a multi-dimensional operation element mining model, and performing element vector mining on the historical operation data to be diagnosed through the multi-dimensional operation element mining model to generate multi-dimensional operation element vectors; taking the X fault trend labels as processing objects of a trend expression element mining model, and performing element vector mining on the X fault trend labels through the trend expression element mining model to generate X trend expression element vectors; invoking a target fault diagnosis algorithm, determining matching coefficients of the multidimensional operation element vector and the X trend expression element vectors through the target fault diagnosis algorithm to obtain X operation fault matching views, wherein the target fault diagnosis algorithm is used for expanding the multidimensional operation element vector and the X trend expression element vectors and determining the matching coefficients of the target multidimensional operation element vector obtained by expansion and the X target trend expression element vectors; and determining a target fault trend label from the X fault trend labels based on the X operation fault matching viewpoints, wherein the operation fault matching viewpoint corresponding to the target fault trend label is the maximum matching viewpoint weight in the X operation fault matching viewpoints.

In a second aspect, the present application further provides an artificial intelligence fault diagnosis system, including: a memory for storing program instructions and data; and a processor coupled to the memory for executing instructions in the memory to implement the method as described above.

In a third aspect, the present application also provides a computer storage medium containing instructions which, when executed on a processor, implement the above-described method.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.

Fig. 1 is a schematic flow chart of a fault diagnosis method of a crankshaft double-top vehicle control system according to an embodiment of the present application.

Fig. 2 is a block diagram of an artificial intelligence fault diagnosis system 300 according to an embodiment of the present application.

Detailed Description

The technical solutions in the present application will be described below with reference to the accompanying drawings.

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with aspects of the present application.

It should be noted that the terms "first," "second," and the like in the description of the present application and the above-described figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.

The method embodiments provided by the embodiments of the present application may be performed in an artificial intelligence fault diagnosis system, a computer device, or similar computing apparatus. Taking as an example operation on an artificial intelligence fault diagnosis system, the artificial intelligence fault diagnosis system may comprise one or more processors (which may include, but is not limited to, a microprocessor MCU or a processing device such as a programmable logic device FPGA) and a memory for storing data, and optionally the artificial intelligence fault diagnosis system may further comprise a transmission device for communication functions. It will be appreciated by those of ordinary skill in the art that the above-described structure is merely illustrative and is not intended to limit the structure of the above-described artificial intelligence fault diagnosis system. For example, the artificial intelligence fault diagnosis system may also include more or fewer components than those shown above, or have a different configuration than those shown above.

The memory may be used to store a computer program, for example, a software program of application software and a module, such as a computer program corresponding to a fault diagnosis method of a crank double-top vehicle control system in the embodiment of the present application, and the processor executes the computer program stored in the memory, thereby performing various functional applications and data processing, that is, implementing the above method. The memory may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid state memory. In some examples, the memory may further include memory remotely located with respect to the processor, the remote memory being connectable to the artificial intelligence fault diagnosis system through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission means is used for receiving or transmitting data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of an artificial intelligence fault diagnosis system. In one example, the transmission means comprises a network adapter (Network Interface Controller, simply referred to as NIC) that can be connected to other network devices via a base station to communicate with the internet. In one example, the transmission device may be a Radio Frequency (RF) module, which is used to communicate with the internet wirelessly.

Referring to fig. 1, fig. 1 is a flow chart of a fault diagnosis method of a crank double-top vehicle control system according to an embodiment of the present application, where the method is applied to an artificial intelligent fault diagnosis system, and further may include steps 110 to 150.

Step 110, obtaining historical operation data to be diagnosed and X fault trend labels, wherein X is an integer not smaller than 1, and the historical operation data to be diagnosed comprises sensor detection data, control output data and actuator state data of a crankshaft double-top vehicle control system.

And 120, taking the historical operation data to be diagnosed as a processing object of a multi-dimensional operation element mining model, and mining element vectors of the historical operation data to be diagnosed through the multi-dimensional operation element mining model to generate multi-dimensional operation element vectors.

And 130, taking the X fault trend labels as processing objects of a trend expression element mining model, and mining element vectors of the X fault trend labels through the trend expression element mining model to generate X trend expression element vectors.

And 140, invoking a target fault diagnosis algorithm, and determining matching coefficients of the multidimensional operation element vector and the X trend expression element vectors through the target fault diagnosis algorithm to obtain X operation fault matching views, wherein the target fault diagnosis algorithm is used for expanding the multidimensional operation element vector and the X trend expression element vectors and determining the matching coefficients of the expanded target multidimensional operation element vector and the X target trend expression element vectors.

And step 150, determining a target fault trend label from the X fault trend labels based on the X operation fault matching views, wherein the operation fault matching view corresponding to the target fault trend label is the maximum matching view weight in the X operation fault matching views.

Firstly, the system extracts historical operation data to be diagnosed from a historical database of a crankshaft double-top vehicle control system. These data include sensor sensing data (e.g., crankshaft position, speed, acceleration, etc.), control output data (e.g., control signals, solenoid status, etc.), and actuator status data (e.g., cylinder position, pressure, etc.). These data reflect the operating state and performance of the control system over time. Meanwhile, the system also acquires X fault trend labels which represent various fault types possibly occurring in the crankshaft double-top vehicle control system, such as sensor faults, actuator faults, communication faults and the like. Each failure trend label corresponds to a particular set of failure characteristics and manifestations.

The system inputs historical operational data to be diagnosed into a multidimensional operational element mining model. This model may be a machine learning model such as a neural network or decision tree. The model performs deep analysis and processing on the data, and extracts key elements, such as a motion track of a crankshaft, a reading change of a sensor and the like. These elements are combined into a multi-dimensional operational element vector that contains the complete information during the operation of the control system.

The system then inputs the X fault trend labels into a trend expression element mining model. This model is similar to the multidimensional operating element mining model, but focuses on extracting key features from fault trend labels. Each failure trend label is converted into a trend expression element vector that describes the typical behavior of the system under different failure trends.

Further, the system invokes a specifically designed target fault diagnosis algorithm. This algorithm first expands the multidimensional running element vector and the X trend-representing element vectors, which may involve operations such as vector dimension increase or feature enhancement. The algorithm then calculates the matching coefficients between the expanded multidimensional running element vector and each trend indicative element vector. This matching factor reflects the degree of similarity between the current operating state of the control system and each fault trend.

Finally, the system determines a target fault trend label according to the calculated matching coefficient. Typically, the failure trend label with the largest matching coefficient will be selected as the target failure trend label. This means that the current state of the control system is closest to the failure trend and thus it is quite likely that a corresponding failure has occurred. This result is very important for subsequent system maintenance, troubleshooting, and preventive maintenance.

In this way, the system can automatically and accurately diagnose the fault type of the crankshaft double-top vehicle control system and provide valuable information to help maintenance personnel to quickly locate and solve the problem.

In step 110, it is referred to a record of operational data generated by the crankshaft dual overhead control system over time, which data is used for subsequent fault diagnosis and analysis. For example, all operational data of a crankshaft dual-top control system over the past week, including time series data of crankshaft position, speed, acceleration, etc., parameters, and status information of the system at various points in time.

The failure trend label is a label for classifying and marking the types of failures which may occur in the crankshaft double-top vehicle control system. Each tag represents a particular failure trend or failure mode. For example, one failure trend label may be "sensor failure" indicating the type of failure associated with the sensor; another label may be "slow actuator response", indicating a tendency for the actuator to slow down in response after receiving a control signal.

The sensor detection data refer to various parameter data collected by sensors in the crankshaft double-top vehicle control system, and the data reflect the running state information such as the position, the speed, the acceleration and the like of the crankshaft. For example, a crankshaft position sensor may generate a series of voltage signals representative of crankshaft angle, which signals vary over time and are recorded as sensor sensed data; a speed sensor may then generate a frequency signal indicative of the rotational speed of the crankshaft.

The control output data refers to a control instruction or a control signal generated by the crankshaft double-top vehicle control system according to the input signal and the internal logic, and is used for driving the actuator to execute corresponding actions. For example, the control system may generate a PWM (pulse width modulation) signal for controlling the opening and closing of the solenoid valve based on the current position and speed of the crankshaft, which is part of the control output data; another example is that the control system outputs a digital signal to control the extension and retraction of the hydraulic cylinder.

The actuator state data refers to data describing the current state of an actuator in the crankshaft double-top vehicle control system, and comprises parameters such as the position, the speed, the pressure and the like of the actuator, and information such as whether the actuator works normally or not, whether faults occur or not and the like. For example, a position sensor of a hydraulic cylinder may generate data indicative of the current position of the hydraulic cylinder, which is part of the actuator status data; a pressure sensor may then generate data indicative of the pressure within the cylinder. In addition, fault conditions of the actuators (e.g., solenoid valve blockage, hydraulic cylinder oil leakage, etc.) may also be recorded as actuator status data.

In steps 120 and 130, the multi-dimensional operating element mining model is a data analysis model for extracting key operating elements from historical operating data of the crankshaft dual overhead vehicle control system and representing the elements in the form of multi-dimensional vectors. This model captures various aspects and features of the control system during operation. For example, the multidimensional operating element mining model is a deep learning model, such as a Convolutional Neural Network (CNN). The motion mode of the crankshaft can be learned from sensor detection data, the control strategy of the system is understood from control output data, and the response characteristic of the actuator is perceived from actuator state data. By integrating this information, the model is able to generate a multi-dimensional running element vector that contains multiple dimensions (e.g., position, velocity, acceleration, control signal strength, etc.).

Element vector mining refers to the process of extracting key operational elements from raw data using a multidimensional operational element mining model or other related algorithm, and converting these elements into vectors. This process aims to convert complex data into a compact and representative mathematical representation. In a crankshaft dual-top control system, element vector mining may involve filtering of sensor detection data to remove noise and extract a true crankshaft motion profile; transcoding the control output data to convert it to a format suitable for model processing; and extracting characteristics of the actuator state data to identify different working states of the actuator.

The multidimensional operating element vector is a vector representation generated by a multidimensional operating element mining model and comprises a plurality of key operating elements of a control system in a historical operating process. Each element corresponds to a dimension in the vector, and by combining these dimensions, the operational state of the system can be comprehensively described. A multi-dimensional running element vector may include the real-time position, speed, acceleration of the crankshaft, and multiple dimensions of the output signal strength of the control system, the response speed of the actuator, etc. Together, these dimensions constitute a comprehensive operational state description of the control system at a certain moment.

The trend expression element mining model is a model specially used for analyzing fault trend labels and extracting key characteristics of the fault trend labels. The model is capable of converting fault trend labels into trend indicative element vectors for comparison and matching with multidimensional operating element vectors. For example, the trending element mining model is a Support Vector Machine (SVM) model. The system performance characteristics under different fault trends can be learned from the historical data, such as abnormal readings when the sensor fails, response delay when the actuator fails, and the like. By learning these features, the model is able to generate a trend indicative element vector that contains multiple dimensions (e.g., fault type, severity, frequency of occurrence, etc.).

The trend-expressing element vector is a vector representation generated by a trend-expressing element mining model that describes the system performance characteristics associated with a particular fault trend. This vector is used to compare with the multidimensional operating element vector to determine whether the control system is currently exhibiting some failure tendency. One trend indicative element vector may include characteristic dimensions such as typical read variability at sensor failure, average response delay time at actuator failure, etc. These dimensions together constitute a comprehensive description of certain fault trends, providing an important basis for subsequent fault diagnosis.

In step 140, the target fault diagnosis algorithm is a specially designed algorithm for diagnosing the fault type of the crank double-top vehicle control system based on the multi-dimensional operation element vector and the trend expression element vector. It makes diagnostic decisions by comparing and analyzing the similarity and relevance of these two classes of vectors. The target fault diagnosis algorithm may be a machine learning based classification algorithm such as a random forest or neural network. It accepts as input the multidimensional running element vector and outputs as a diagnostic result the fault type corresponding to the most matching trend indicative element vector.

The matching coefficient is a quantization index used for measuring the similarity or the association degree between the multidimensional operation element vector and the trend expression element vector. A higher matching factor means that the two vectors are more similar in a sense, possibly indicating a corresponding failure trend. The matching coefficient can be cosine similarity between two vectors, and the value range is between-1 and 1. When the cosine similarity approaches 1, it means that the two vectors are very similar; approaching-1 indicates that they are very different.

The operation failure matching viewpoint refers to a viewpoint or judgment formed by matching coefficients between the multidimensional operation element vector and each trend expression element vector calculated by the target failure diagnosis algorithm. Each matching view reflects the matching degree of the current running state of the control system and a certain fault trend. If there are three failure trend labels A, B and C, the algorithm may calculate three matching coefficients, corresponding to the three labels, respectively. The three matching coefficients form three operational fault matching views, which indicate the similarity of the current state of the control system and A, B, C three fault trends.

In this context, expansion refers to the process of enhancing or transforming multidimensional running element vectors and trend performance element vectors to increase their dimensions or extract more meaningful features, thereby improving the accuracy of fault diagnosis. The expansion may include operations such as normalization of the vector, feature scaling, principal Component Analysis (PCA) dimension reduction, or feature selection. These operations may help highlight important features in the vector and eliminate redundant information.

The target multidimensional operation element vector refers to the multidimensional operation element vector after the expansion processing. It contains richer, more meaningful running element information for comparison and matching with the extended trend indicative element vector. For example, the original multidimensional operation element vector contains the position and speed information of the crankshaft, and after the expansion processing, the target multidimensional operation element vector may further comprise additional characteristics such as acceleration, vibration frequency and the like.

The target trend expression element vector is a trend expression element vector after the expansion processing. It contains richer, more meaningful feature information related to specific fault trends for comparison and matching with the target multidimensional operating element vector. For example, the raw trend indicative element vector describes typical readings of the sensor fault, and after the expansion process, the target trend indicative element vector may also include other relevant characteristics of the sensor fault, such as the frequency or duration of occurrence of the fault.

The matching coefficient determination refers to a process of calculating a matching coefficient between the target multidimensional operating element vector and each target trend representing element vector using a target fault diagnosis algorithm. This process aims to quantify the degree of similarity or correlation between the current operating state of the control system and various fault trends. In making the matching coefficient determination, the algorithm may calculate cosine similarity, euclidean distance, or other relevant metric between the target multidimensional operating element vector and each target trend representative element vector, and derive a matching coefficient matrix based on these metrics, where each element represents an operating fault matching perspective.

In step 150, the target fault trend label refers to the fault trend label that is determined to be the best match with the current system running state after the multidimensional running element vector is analyzed by the target fault diagnosis algorithm. This tag represents the type or trend of failure that the system may have. For example, a certain abnormality occurs in the operation process of the crankshaft double-top vehicle control system, and after the multidimensional operation element vector is analyzed through the target fault diagnosis algorithm, a target fault trend label is determined as 'sensor drift'. This tag indicates that the system may be currently experiencing the problem of the sensor reading gradually deviating from the true value.

The maximum matching viewpoint weight is a weight corresponding to the viewpoint having the highest matching coefficient among the plurality of operation failure matching viewpoints. This weight reflects the importance or impact of the matching perspective in the fault diagnosis decision. For example, three operational fault matching points of view correspond to three fault trends of "sensor fault", "actuator response delay", and "control logic error", respectively, and their matching coefficients are 0.9, 0.5, and 0.3, respectively. In this case, the maximum matching viewpoint weight is the weight corresponding to the matching viewpoint of "sensor malfunction" because it has the highest matching coefficient of 0.9. This maximum matching perspective weight may be used to strengthen or highlight the importance of the failure trend of "sensor failure" during subsequent failure diagnosis.

The application provides a crankshaft double-top-vehicle control system fault diagnosis method based on multidimensional operation element mining and trend expression element mining, which has the following advantages:

comprehensive data acquisition: through step 110, the method and the device not only acquire the sensor detection data and the control output data of the crankshaft double-top vehicle control system, but also consider the state data of the actuator, thereby ensuring the comprehensiveness and the accuracy of the historical operation data to be diagnosed. The comprehensive data acquisition mode provides a solid foundation for subsequent fault diagnosis;

efficient element vector mining: step 120 and step 130 perform element vector mining on historical operating data to be diagnosed and fault trend labels by using a multidimensional operating element mining model and a trend expression element mining model. The mining method can effectively extract key features in the data and convert the key features into a vector form, so that subsequent processing and analysis are facilitated;

accurate fault matching: step 140 performs matching coefficient determination on the multidimensional operation element vector and the trend expression element vector through a target fault diagnosis algorithm. The algorithm firstly expands the vector, further enriches the characteristic information of the vector, then calculates the matching coefficient, and obtains a plurality of operation fault matching views. The accurate matching mode ensures the accuracy of fault diagnosis;

Accurate target fault trend label determination: step 150 determines a target failure trend label from a plurality of failure trend labels based on an operational failure matching standpoint. The determining mode fully considers the influence of each matching view, especially the influence of the maximum matching view weight, so that the accuracy and the reliability of the target fault trend label are ensured.

In summary, the method and the device realize efficient and accurate fault diagnosis of the crankshaft double-top vehicle control system through comprehensive data acquisition, efficient element vector mining, accurate fault matching and accurate target fault trend label determination. The diagnosis method not only can timely find and locate the faults of the system, but also can provide powerful support for repairing and preventing the faults, thereby improving the stability and reliability of the system.

Under some exemplary design considerations, a target fault diagnosis algorithm is invoked in step 140, and the matching coefficients of the multidimensional operation element vector and the X trend expression element vectors are determined by the target fault diagnosis algorithm, so as to obtain X operation fault matching perspectives, including steps 141-144.

And 141, determining a target knowledge relation network corresponding to the target fault diagnosis algorithm based on the target fault diagnosis algorithm.

And 142, projecting the multidimensional operation element vector to the target knowledge relation network to obtain the target multidimensional operation element vector, wherein the characteristic size of the target multidimensional operation element vector is larger than that of the multidimensional operation element vector.

And step 143, projecting the X trend expression element vectors to the target knowledge relationship network to obtain X target trend expression element vectors, where the feature size of the target trend expression element vectors is greater than the feature size of the trend expression element vectors.

And 144, invoking the target fault diagnosis algorithm to determine matching coefficients of the target multidimensional operation element vector and the X target trend expression element vectors, so as to obtain X operation fault matching views.

In this step, the system processes the multidimensional running element vector and the trend expression element vector using a predefined target fault diagnosis algorithm. The core objective of the algorithm is to determine the matching coefficients between these two classes of vectors, resulting in multiple operational fault matching perspectives.

Firstly, the system needs to determine a target knowledge relation network corresponding to a target fault diagnosis algorithm. The knowledge relation network is a complex network structure and comprises various fault modes, fault symptoms, fault reasons and correlations among the fault modes, the fault symptoms and the fault reasons which can be encountered in the operation process of the crankshaft double-top-vehicle control system. The knowledge can be obtained from a plurality of sources such as historical fault data, expert experience, technical literature and the like, and is processed and analyzed to form a structured knowledge network.

For example, the target knowledge relationship network may be a graph structure composed of nodes and edges, wherein the nodes represent different failure modes, symptoms and causes, and the edges represent the association relationships between them. Such a graph structure may effectively represent the complexity and relevance between faults.

Next, the system projects the multi-dimensional running element vector onto a target knowledge relationship network to obtain a target multi-dimensional running element vector. This process can be seen as mapping the original, possibly sparse, running element vector into a denser and high-dimensional feature space.

For example, if the multidimensional operating element vector originally contained only features of the sensor detection data, it could be extended to a target multidimensional operating element vector containing more fault-related features (e.g., fault patterns, symptoms, etc.) by projection onto a target knowledge-relationship network. In this way, the system can describe and analyze the current operating conditions from a more comprehensive perspective.

Similarly, the system also needs to project the X trend-expressing element vectors onto the target knowledge relationship network to obtain X target trend-expressing element vectors. This process is similar to step 142, but the object processed is a trend expression element vector.

The trend indicative element vector may also be extended into a higher dimensional feature space by projection, thereby containing more information about failure trends. In this way, the system can more accurately identify and describe different fault trends.

And finally, the system calls a target fault diagnosis algorithm to determine matching coefficients of the target multidimensional operation element vector and the X target trend expression element vectors. This process typically involves complex mathematical calculations and pattern recognition techniques.

For example, the system may calculate the similarity or difference between the two classes of vectors using a measurement such as cosine similarity, euclidean distance, etc., to obtain the matching coefficients between them. The matching coefficients can intuitively reflect the matching degree between the current running state and different fault trends.

Through this step, the system finally obtains X operational fault matching views, each view corresponding to a fault trend label and a matching coefficient. These matching perspectives can provide an important reference basis for subsequent fault diagnosis and processing.

In some preferred embodiments, the projecting the multidimensional operating element vector into the target knowledge-relation network as described in step 142 results in the target multidimensional operating element vector, including steps 1421-1424.

Step 1421, obtaining a knowledge vector derivative model.

Step 1422, determining the feature size of the target knowledge-relation network.

Step 1423, optimizing the variables of the knowledge vector derivative model based on the feature size of the target knowledge relationship network to obtain a target knowledge vector derivative model corresponding to the target knowledge relationship network.

Step 1424, using the multidimensional operation element vector as a processing object of the target knowledge vector derivative model, and expanding the multidimensional operation element vector through the target knowledge vector derivative model to obtain the target multidimensional operation element vector, wherein the feature size of the target multidimensional operation element vector is consistent with the feature size of the target knowledge relationship network.

In order to deal with the fault diagnosis of the crank double-top vehicle control system, the system needs to acquire a key tool, namely a knowledge vector derivative model. This model is an algorithm specifically designed to extract and derive deep knowledge related to fault diagnosis from raw data. The system can be assisted to convert the multidimensional operation element vector into a richer and more expressive form.

Next, the system determines a feature size of the target knowledge-relationship network. Feature size is an important parameter that defines the dimensions and complexity of information in a knowledge relationship network. By knowing the feature size of the target knowledge relationship network, the system can ensure that subsequent data processing and analysis work can be performed on the correct scale.

Next, the system optimizes variables of the knowledge vector derivative model based on feature sizes of the target knowledge-relationship network. This optimization process is to ensure that the knowledge vector derivation model is better able to adapt to the structure and features of the target knowledge-relationship network. By adjusting the variables of the model, the system can enable the derived knowledge vector to be more matched with the target knowledge relation network, so that the accuracy and the efficiency of fault diagnosis are improved.

And finally, the system takes the multidimensional operation element vector as a processing object of the optimized target knowledge vector derivative model. Through the model, the system expands the multidimensional operation element vector to obtain a target multidimensional operation element vector. This expansion process can be seen as an enhancement or dimension-up operation on the raw data that causes the multidimensional running element vector to contain more features and information about the target knowledge relationship network. Thus, the system can more fully describe and analyze the operation state of the crankshaft double-top-car control system, and thus the potential faults can be diagnosed more accurately.

In this process, it is critical that the feature size of the target multidimensional operating element vector be consistent with the feature size of the target knowledge relationship network. The consistency of the information expression of the system and the target knowledge relationship network is ensured, so that the system can more effectively utilize the knowledge in the target knowledge relationship network to perform fault diagnosis. In this way, the system not only improves the accuracy of fault diagnosis, but also enhances its adaptability and recognition of new fault types.

In addition, the knowledge vector derivation model can be applied to different types of neural networks. The core idea of knowledge vector derivation models is to extract and derive deep knowledge related to specific tasks (such as fault diagnosis) from raw data, which is usually expressed in the form of vectors and can be used to enhance the performance and generalization ability of neural networks.

In the example of a specific neural network, the relevant concepts of the knowledge vector derivation model may be introduced in consideration of the application of Convolutional Neural Networks (CNNs) in image processing. While CNN is not itself a knowledge vector derived model, some of its features and components may be used to construct such a model.

Convolution layer and feature extraction: in CNN, the convolution layer is responsible for extracting local features from the input image. These features can be seen as knowledge vectors in the image, which describe important information of different areas in the image. By stacking multiple convolution layers, the CNN is able to progressively extract higher-level feature representations that are very useful for subsequent classification or regression tasks.

Full connectivity layer and knowledge integration: at the end of the CNN there are typically one or more fully connected layers for integrating and mapping the features extracted by the convolutional layers to the final output space. These fully connected layers can be considered as part of the knowledge vector derivation model in that they combine and transform different levels of features to produce outputs related to specific tasks.

In this example, the knowledge vector derivation model can be built by adding additional layers or modules on top of the CNN. For example, a self-encoder (auto-encoder) structure may be introduced before the fully-connected layer for learning a compressed representation (i.e., knowledge vector) of the input data. This compressed representation may capture critical information in the input data and be used in subsequent tasks to enhance the performance of the neural network.

In addition, the knowledge vector derivative model can also be used for processing the input data with time sequence relation by referencing the ideas of other neural network architectures, such as the sequence modeling capability in a cyclic neural network (RNN). By combining these neural network components with the concept of knowledge vectors, a more powerful, flexible neural network model can be built to cope with a variety of complex tasks and data types.

Under other optional design considerations, the invoking the target fault diagnosis algorithm in step 144 determines a matching coefficient of the target multidimensional operation element vector and the X target trend expression element vectors, to obtain X operation fault matching perspectives, including: acquiring the characteristic size of the target multidimensional operation element vector and the characteristic size of each target trend expression element vector in the X target trend expression element vectors; and carrying out operation fault decision on the target multidimensional operation element vector and each target trend expression element vector based on the characteristic size of the target multidimensional operation element vector and the characteristic size of each target trend expression element vector in the X target trend expression element vectors to obtain X operation fault matching views.

In the fault diagnosis process of the crankshaft double-top vehicle control system, when the system needs to determine the matching degree between the multidimensional operation element vector and the trend expression element vector, a refined processing method is adopted. The method not only considers the numerical information of the vectors, but also deeply analyzes the characteristic sizes of the vectors, thereby improving the accuracy and the reliability of fault diagnosis.

First, the system obtains the feature size of the target multidimensional operating element vector. Feature size is the dimension or length of each feature in a vector, which reflects the distribution and importance of the feature in vector space. By knowing the feature size, the system can more accurately grasp the overall structure and key information of the multidimensional operation element vector.

Second, for each of the X target trend indicative element vectors, the system will also obtain its corresponding feature size. These trend-expressing element vectors represent different fault trends or patterns, and their feature sizes reveal the uniqueness and discrimination of these trends in vector space.

Next, the system makes an operational fault decision based on the feature size of the target multidimensional operational element vector and the feature size of each target trending element vector. The process involves complex mathematical calculations and pattern recognition techniques, and the system compares the similarity or difference in feature sizes of the two types of vectors to determine how well the current operating state matches each fault trend.

And finally, the system obtains X operation fault matching views according to the operation fault decision result. Each view corresponds to a fault trend and gives a matching coefficient that quantifies the similarity of the current operating state to the fault trend. These matching perspectives provide powerful support for subsequent fault diagnosis and handling of the system, helping the system to locate and solve potential fault problems more quickly and accurately.

By the method, the system can fully utilize information in the multidimensional operation element vector and the trend expression element vector, and realize comprehensive monitoring and accurate diagnosis of the operation state of the crankshaft double-top vehicle control system.

In other possible application scenarios, first, the system has acquired a target multidimensional running element vector and X target trend performance element vectors. These vectors contain various key information during system operation, such as sensor readings, operating parameters, etc.

Next, the system will perform an inner product operation on the target multidimensional run element vector and each target trend expression element vector. Specifically, the system calculates the sum of the products of the corresponding elements of the two vectors to obtain a scalar value. This scalar value reflects the similarity in value and direction of the target multidimensional run element vector to the current trend indicative element vector.

When performing the inner product operation, the system may perform normalization processing on the vector first to eliminate the influence of the vector length on the result. The normalization process may scale the length of the vector to a unit length, thereby making the inner product operation more focused on differences in the vector direction.

After the inner product operation is completed, the system obtains a scalar value as a matching coefficient. This matching coefficient quantifies the similarity of the target multidimensional operating element vector to the current trend indicative element vector. The larger the matching coefficient is, the more similar the current running state and the fault trend are; otherwise, the similarity is lower.

The system performs the inner product operation described above on each of the target multidimensional run element vector and the X target trend indicative element vectors to obtain X matching coefficients. The matching coefficients form X operation fault matching views, each view corresponds to a fault trend, and the similarity between the current operation state and the fault trend is given.

Finally, the system may make fault decisions based on these matching coefficients. For example, the system may set a threshold value, and when a certain matching coefficient exceeds the threshold value, the current running state is considered to be highly matched with the fault trend, so as to send out corresponding fault early warning or take other processing measures.

By taking inner product operation as an example, the system can realize accurate monitoring and fault diagnosis of the running state of the crankshaft double-top vehicle control system. The scheme fully utilizes the information in the multidimensional operation element vector and the trend expression element vector, and judges the matching degree of the current operation state and each fault trend by calculating the similarity between the vectors, thereby providing powerful support for subsequent fault processing.

In some exemplary embodiments, the step 120 uses the historical operation data to be diagnosed as a processing object of a multi-dimensional operation element mining model, and the multi-dimensional operation element mining model is used for element vector mining on the historical operation data to be diagnosed to generate a multi-dimensional operation element vector, which includes steps 121-123.

And 121, taking the historical operation data to be diagnosed as a processing object of a data disassembly sub-model in the multi-dimensional operation element mining model, and disassembling the historical operation data to be diagnosed through the data disassembly sub-model to obtain Y target operation data blocks, wherein Y is an integer greater than 1.

And 122, taking the Y target operation data blocks as processing objects of a distribution detection sub-model in the multi-dimensional operation element mining model, and determining the distribution characteristics of each target operation data block in the Y target operation data blocks through the distribution detection sub-model to obtain Y operation data block distribution characteristics.

And 123, taking the Y target operation data blocks and the Y operation data block distribution characteristics as processing objects of a characteristic focusing sub-model in the multi-dimensional operation element mining model, and performing element extraction on the Y target operation data blocks and the Y operation data block distribution characteristics through the characteristic focusing sub-model to generate a multi-dimensional operation element vector.

In the fault diagnosis process of the crankshaft double-top vehicle control system, the system needs to process a large amount of historical operation data so as to mine out multidimensional operation element vectors. These vectors are key inputs for subsequent fault diagnosis. To achieve this goal, the system employs a multidimensional running element mining model and generates the required multidimensional running element vector through a series of refined processing steps.

Firstly, the system takes historical operation data to be diagnosed as a processing object of a data disassembly sub-model in the multi-dimensional operation element mining model. The design purpose of the data disassembly sub-model is to disassemble complex historical operating data into smaller, more manageable data blocks. By processing the data disassembly sub-model, the system will obtain Y target operational data blocks, where Y is an integer greater than 1. Each target operational data block contains a portion of the information in the original historical operational data and these data blocks will be analyzed as separate units in subsequent processing.

Second, the system will take these Y target operational data blocks as the processing objects of the distribution detection sub-model in the multidimensional operational element mining model. The main task of the distribution detection sub-model is to determine the distribution characteristics of each target operational data block. These distribution features describe the statistical regularity of the data in the data block, such as central tendency, degree of dispersion, etc. By processing the distribution detection sub-model, the system obtains Y distribution characteristics of the operation data blocks, wherein each characteristic corresponds to one target operation data block.

Finally, the system takes the Y target operation data blocks and the corresponding Y operation data block distribution characteristics as the processing objects of the characteristic focusing sub-model in the multi-dimensional operation element mining model. The feature focus sub-model is a core part of the whole mining model, and is used for element extraction of the target operation data block and the distribution features thereof. Element refinement is a complex process that involves in-depth analysis and understanding of the data to extract the most representative features. Through the processing of the feature focus sub-model, the system ultimately generates a multi-dimensional running element vector. The vector contains information of multidimensional operation elements mined from historical operation data, and is an important basis for subsequent fault diagnosis.

Through the steps, the system can realize deep excavation and analysis of historical operation data of the crankshaft double-top vehicle control system, and a multidimensional operation element vector is generated. These vectors not only contain rich operational information, but also provide powerful support for subsequent fault diagnosis.

Further, the data disassembly sub-model, the distribution detection sub-model and the feature focus sub-model are three key components in the multi-dimensional operational element mining model, which play an important role in processing and analyzing historical operational data. These three sub-models will be described below by way of specific examples of algorithms.

The main task of the data disassembly sub-model is to disassemble complex historical operating data into smaller, more manageable data blocks. This may be achieved by a variety of algorithms, such as a clustering algorithm or a segmentation algorithm.

Clustering algorithm example: k-means clustering is a commonly used method of data disassembly. The system may treat the historical operating data as a set of points in a high-dimensional space, and then use the K-means algorithm to divide the points into Y clusters. Each cluster represents a target operational data block, which allows the raw data to be broken down into smaller portions.

The distribution detection sub-model is used to determine the distribution characteristics of each target operational data block. This typically involves statistical analysis and probability distribution fitting.

Statistical analysis example: for each target operational data block, the system may calculate statistics of its mean, variance, skewness, kurtosis, etc., which may describe the distribution characteristics of the data. In addition, the system can also visualize the distribution of data by plotting a histogram or a kernel density estimation map.

Probability distribution fitting example: the system may attempt to fit the data for each target operational data block using a different probability distribution (e.g., normal distribution, exponential distribution, poisson distribution, etc.), and select the best fit distribution as the distribution characteristics for that data block.

The feature focus sub-model is a core part of the whole mining model, and is used for element extraction of the target operation data block and the distribution features thereof. This typically involves feature selection and feature transformation.

Feature selection example: the system may use statistical-based methods (e.g., analysis of variance, chi-square test) or machine-learning-based methods (e.g., decision trees, feature importance assessment of random forests) to select the most representative features in each target operational data block.

Feature transformation example: for selected features, the system may be further transformed to enhance its expressivity. Common feature transformation methods include normalization, discretization, logarithmic transformation, and the like. In addition, the system can also perform dimension reduction processing through a Principal Component Analysis (PCA) method, a Linear Discriminant Analysis (LDA) method and the like so as to extract the principal components of the data.

In summary, the data disassembly sub-model, the distribution detection sub-model and the feature focus sub-model together form a core framework of the multi-dimensional operation element mining model. Through the synergistic effect of the submodels, the system can realize deep mining and analysis of historical operation data, extract multidimensional operation element vectors and provide powerful support for subsequent fault diagnosis.

In other possible embodiments, the step 123 uses the Y target operation data blocks and the Y operation data block distribution features as the processing objects of the feature focusing sub-model in the multi-dimensional operation element mining model, and performs element refinement on the Y target operation data blocks and the Y operation data block distribution features through the feature focusing sub-model to generate a multi-dimensional operation element vector, which includes steps 1231-1233.

Step 1231, using the Y target operation data blocks as processing objects of feature element focusing nodes in the feature focusing sub-model, and determining matching coefficients of every two target operation data blocks in the Y target operation data blocks through the feature element focusing nodes to obtain Z data block matching coefficients, where z=y (Y-1)/2, and Z is an integer greater than 1.

And 1232, taking the Y target operation data blocks as processing objects of feature embedding nodes in the feature focusing sub-model, and extracting element knowledge of each target operation data block in the Y target operation data blocks through the feature embedding nodes to obtain element knowledge of the Y operation data blocks.

Step 1233, using the Z data block matching coefficients, the element knowledge of the Y operation data blocks, and the distribution features of the Y operation data blocks as processing objects of feature integration nodes in the feature focusing sub-model, and performing element integration on the Z data block matching coefficients, the element knowledge of the Y operation data blocks, and the distribution features of the Y operation data blocks through the feature integration nodes to obtain a multidimensional operation element vector.

In order to extract a multidimensional operating element vector from a large amount of historical operating data in the fault diagnosis process of the crankshaft double-top vehicle control system, the system utilizes a multidimensional operating element mining model. In the final stage of this model, the feature focusing sub-model plays a key role and is responsible for deep element extraction of the Y target operation data blocks and the distributed features thereof obtained in the previous step.

First, the system inputs Y target operational data blocks into feature element focus nodes in a feature focus sub-model. The design of the feature element focus node is aimed at evaluating the similarity or correlation between these data blocks. To achieve this, the system calculates the matching coefficients between every two target operational data blocks. These matching coefficients reflect how similar the data blocks are in value, distribution, or other relevant characteristics. By this step, the system will get Z data block matching coefficients, where Z is equal to Y times (Y minus 1) divided by 2, ensuring that each pair of data blocks is uniquely compared once.

Second, the system will embed the same Y target operational data blocks as the processing objects of the feature embedding nodes in the feature focusing sub-model. The task of the feature embedding node is to mine the knowledge of the elements inside each data block deeply. These element knowledge may be specific patterns in the data block, outliers, periodic behavior, or other information that is valuable for fault diagnosis. By processing the feature embedding nodes, the system can extract key features of each target operation data block to obtain element knowledge of Y operation data blocks.

Finally, the system inputs the Z data block matching coefficients, Y pieces of operation data block element knowledge and Y pieces of operation data block distribution characteristics into the characteristic integration nodes in the characteristic focusing submodel. The feature integration node is used for comprehensively integrating the information from different layers. In this process, the system may use various machine learning algorithms or statistical techniques to ensure the validity and accuracy of the integration. The result of the integration is a multidimensional operating element vector that contains not only the key information in the original historical operating data, but also is further refined and optimized by the processing of the feature focused sub-model.

Through the steps, the system can realize comprehensive excavation and deep analysis of historical operation data of the crankshaft double-top vehicle control system, and generate multidimensional operation element vectors which are vital to fault diagnosis.

In the multidimensional operating element mining model, feature element focusing nodes, feature embedding nodes and feature integration nodes are key components in a feature focusing sub-model. These three nodes will be described below by way of specific examples of algorithms.

The goal of the feature element focus node is to determine the matching coefficients between different target operational data blocks to quantify the similarity or correlation between them.

Example of algorithm: cosine similarity is a commonly used method for calculating the similarity between vectors, and can be used for focusing nodes of characteristic elements. The system first represents each target operational data block as a feature vector and then calculates the cosine similarity between each pair of feature vectors. The cosine similarity has a value between-1 and 1, with a value closer to 1 indicating a higher similarity between the vectors. In this way, the system can obtain a matrix of matching coefficients, where each element represents the similarity between a pair of target operational data blocks.

The task of the feature embedding node is to extract the knowledge of the elements in each target operational data block, i.e. the features or patterns inside the data block.

Example of algorithm: self-encoder is a common method of unsupervised feature learning that can be used for feature embedding nodes. The system first builds a self-encoder neural network, which is composed of two parts, an encoder and a decoder. The encoder compresses the incoming target operational data block into a low-dimensional hidden representation, and the decoder attempts to reconstruct the original data block from the hidden representation. By training the self-encoder to minimize reconstruction errors, the system can learn useful features in the data block, which are stored in the hidden layer of the encoder as element knowledge.

The goal of the feature integration node is to integrate information from the feature element focus node and the feature embedding node comprehensively to generate a multidimensional running element vector.

Example of algorithm: the integration method may employ simple stitching (registration) or more complex fusion strategies such as weighted averaging, neural network fusion, etc. Taking the splicing as an example, the system can splice the matching coefficient matrix, the element knowledge vector and other related features to form a longer feature vector. The feature vector is a multidimensional operation element vector, integrates information of different layers, and can be used for subsequent fault diagnosis or other related tasks.

It should be noted that the above example of algorithm is only one possible implementation, and that in practical applications, suitable algorithms and techniques may be selected according to specific requirements and data characteristics. Furthermore, the design and implementation of the feature focus sub-model may involve more detail and optimization to ensure that the extracted multidimensional running element vector has better representation capabilities and diagnostic performance.

In still other exemplary embodiments, in step 121, the step of taking the historical operation data to be diagnosed as a processing object of a data disassembly sub-model in the multi-dimensional operation element mining model, and the step of disassembling the historical operation data to be diagnosed through the data disassembly sub-model, to obtain Y target operation data blocks, includes steps 1211 to 1213.

And 1211, taking the historical operation data to be diagnosed as a processing object of a dismantling node in the data dismantling sub-model, and dismantling the historical operation data to be diagnosed through the dismantling node to generate Y operation data units.

And 1212, taking the Y operation data units as processing objects of feature coding nodes in the data disassembly sub-model, and performing feature coding on the Y operation data units through the feature coding nodes to generate Y operation data block coding sets.

And 1213, using the Y running data block code sets as processing objects of data updating nodes in the data disassembly sub-model, and performing data updating on the Y running data block code sets through the data updating nodes to generate Y target running data blocks, wherein the feature size corresponding to the target running data blocks is smaller than the feature size of the running data unit.

In the fault diagnosis process of the crankshaft double-top vehicle control system, when the system needs to take the historical operation data to be diagnosed as the input of the multi-dimensional operation element mining model, the function of data disassembly sub-model is particularly critical. This sub-model is responsible for breaking down complex and huge historical operating data into smaller, more tractable data blocks, providing the basis for subsequent distribution detection and feature focusing.

First, the system will input historical operational data to be diagnosed into the disassembled nodes in the data disassembly sub-model. The working mechanism of the node disassembly is similar to an advanced data segmentation tool, and the node disassembly is performed on historical operation data to be diagnosed according to preset rules or algorithms, such as window sliding of time sequences, triggering conditions of events and the like. This disassembly is to break the continuity or complexity of the original data, making it easier to analyze and process. After node disassembly, the system generates Y units of operational data, each unit containing a portion of the information in the original data.

Second, these units of operational data are passed further to feature encoding nodes in the data de-modeling sub-model. The function of the feature encoding node is to perform feature level conversion or extraction on each operation data unit. Such conversion may be to encode certain features in the original data, such as converting a textual description to a digital code, or to extract key statistical features in the data, etc. Each of the operation data units is converted into a corresponding operation data block code set by the processing of the feature code node, and the code set contains the expression of the data unit at the feature level.

Finally, the running data block code sets are input to the data update nodes in the data de-modeling sub-model. The data update node works to further optimize or filter the code set to ensure that each target operational data block contains the most valuable information. Such updating may involve dimension reduction of the data, screening or recalculation of features, etc. After processing by the data update node, the system generates Y target operational data blocks. These target operational data blocks not only contain critical information in the original historical operational data, but also typically have smaller feature sizes than the original operational data units, which makes subsequent processing and analysis more efficient and accurate.

Through the steps, the system can effectively disassemble and process the historical operation data to be diagnosed by utilizing the data disassembly sub-model, and the generated target operation data block provides a solid foundation for subsequent multi-dimensional operation element excavation.

In some possible embodiments, the step 130 uses the X fault trend labels as a processing object of a trend expression element mining model, and the element vector mining is performed on the X fault trend labels by using the trend expression element mining model, so as to generate X trend expression element vectors, which includes steps 131-135.

And 131, taking each fault trend label in the X fault trend labels as a processing object of a stage trend disassembling node in the trend expression element mining model, and carrying out stage trend disassembling on each fault trend label in the X fault trend labels through the stage trend disassembling node to obtain the X stage trend sub-label sets.

And 132, using the X stage trend sub-label sets as processing objects of trend vector recognition sub-models in the trend expression element mining model, and processing each stage trend sub-label set in the X stage trend sub-label sets through the trend vector recognition sub-models to generate X stage trend characterization vector sets.

And 133, using the X stage trend characterization vector sets as processing objects of stage trend attention sub-models in the trend expression element mining model, and processing each stage trend characterization vector set in the X stage trend characterization vector sets through the stage trend attention sub-models to generate X stage trend attention vector sets, wherein the stage trend attention vector sets comprise at least one stage trend attention vector.

Step 134, integrating the stage trend attention vectors in each stage trend attention vector set in the X stage trend attention vector sets to obtain X fault trend integrated vectors.

And 135, performing interval numerical mapping on each fault trend integration vector in the X fault trend integration vectors to obtain X trend expression element vectors.

In the fault diagnosis flow of the crankshaft double-top vehicle control system, when the system needs to deeply excavate element vectors behind fault trend labels, a trend expression element excavation model plays a key role. The model can generate trend expression element vectors with rich expressive force through careful processing of fault trend labels.

First, the system will input the X fault trend labels one by one to the stage trend dismantling nodes in the trend expression element mining model. The task of this node is to split the phase trend for each fault trend label, i.e. subdivide the fault trend contained in the label according to different phases or characteristics. After disassembly, the system obtains X stage trend sub-label sets, each set containing the performance of the corresponding fault trend label at different stages or levels.

Second, these stage trend sub-label sets are fed into a trend vector recognition sub-model for processing. The purpose of this sub-model is to identify and extract features in each stage trend sub-label set by a specific algorithm or model, such as a Recurrent Neural Network (RNN) or long short term memory network (LSTM) in deep learning, and thus generate a corresponding stage trend characterization vector set. These token vector sets numerically characterize key features of the stage trend sub-label set.

The system then takes these sets of stage trend token vectors as inputs to the stage trend attention sub-model. The task of this sub-model is to screen out the most representative trend information in each stage by further analysis and processing of the token vector and to generate a corresponding set of stage trend attention vectors. These sets of attention vectors not only reduce the information, but also highlight the core trend features in each stage.

The system then integrates the attention vectors in each stage trend attention vector set. This step typically involves combining multiple vectors of interest into one more globally representative failure trend integration vector through weighted averaging, concatenation, or other integration strategies. Thus, each fault trend label corresponds to an integrated vector integrating key trend information of each stage.

Finally, the system maps the interval values of the fault trend integrated vectors. This step is to convert the range of values of the integrated vector to a uniform scale for subsequent analysis and comparison. After mapping, the system finally obtains X trend expression element vectors, and the vectors not only integrate the information of each stage of the fault trend label, but also have a unified numerical expression form, thereby providing powerful data support for subsequent fault diagnosis.

In other possible embodiments, the processing the X stage trend characterization vector sets in step 133 as the processing object of the stage trend focus sub-model in the trend expression element mining model, processing each stage trend characterization vector set in the X stage trend characterization vector sets by the stage trend focus sub-model, to generate X stage trend focus vector sets, including: each stage trend representation vector set in the X stage trend representation vector sets is used as a processing object of a characteristic focusing node in the stage trend focusing sub-model, semantics among the stage trend representation vectors in each stage trend representation vector set are identified through the characteristic focusing node, and X first fault representation focusing vector sets are obtained, wherein the first fault representation focusing vector sets comprise at least one first fault representation focusing vector; taking each first failure expression focusing vector set in the X first failure expression focusing vector sets as a processing object of a first fusion node in the stage trend focusing sub-model, and carrying out integrated processing on the first failure expression focusing vector in each first failure expression focusing vector set through the first fusion node to obtain X first fusion vectors; taking the X first fusion vectors as a processing object of a first interval value mapping node in the stage trend concern submodel, and carrying out interval value mapping on the X first fusion vectors through the first interval value mapping node to obtain X first quantized trend vectors; and taking the X first quantized trend vectors as processing objects of feature embedding nodes in the stage trend attention sub-model, and identifying semantic information between every two first quantized trend vectors in the X first quantized trend vectors through the feature embedding nodes to obtain X stage trend attention vector sets, wherein each stage trend attention vector set comprises X (X-1)/2 stage trend attention vectors.

In the fault diagnosis process of the crankshaft double-top vehicle control system, when the system needs to deeply process the phase trend characterization vector set, the phase trend attention sub-model in the trend expression element mining model plays an important role. This sub-model can further refine the key trend attention vector by fine processing of the token vector.

First, the system inputs a set of X stage trend characterization vectors to feature focus nodes in a stage trend attention sub-model. The core task of the feature focus node is to identify semantic relationships between the vectors within each stage trend token vector set. Such semantic relationships may be manifested in similarities, differences between vectors, or in a particular pattern of associations. Through the processing of the characteristic focusing nodes, the system can extract the most representative fault trend information in each set, and then generate X first fault representation focusing vector sets. Each such set contains at least one first fault-exhibiting focus vector that numerically characterizes key fault signatures in the corresponding set of stage trend characterization vectors.

Secondly, the system transmits the X first failure expression focusing vector sets to a first fusion node for processing. The first fusion node functions to integrate the focus vectors in each set. Such integration may involve weighted averaging, concatenation or other complex mathematical operations with the aim of combining multiple focus vectors into one first fusion vector that is more globally representative. Through the processing of the first fusion nodes, the system obtains X first fusion vectors, and each vector synthesizes a plurality of key fault characteristics in the corresponding stage trend characterization vector set.

These first fusion vectors are then input to the first interval value mapping node for further processing. The task of this node is to map the range of values of each first fusion vector into a uniform interval. The mapping is helpful to eliminate analysis difficulty caused by dimension or numerical range differences among different vectors, so that subsequent processing is more convenient and accurate. Through the processing of the first interval value mapping node, the system obtains X first quantization trend vectors, and the vectors are comparable in value.

Finally, the system takes the X first quantization trend vectors as inputs to the feature embedding node. The goal of the feature embedding node is to identify semantic information between every two first quantized trend vectors. Such semantic information may be embodied in dependencies between vectors, dependencies, or on some particular structural feature. Through the processing of feature embedding nodes, the system is able to generate a set of X stage trend attention vectors. Each set contains X (X-1)/2 phase trend attention vectors, which numerically describe key semantic relationships among different first quantization trend vectors, and provide abundant data support for subsequent fault diagnosis.

In some independent embodiments, the method further comprises steps 210-250.

Step 210, acquiring an operation data sample and a fault trend sample, wherein the fault trend sample is used for reflecting fault trend information of the operation data sample.

And 220, taking the operation data sample as a processing object of a multi-dimensional operation element mining model, and performing element vector mining on the operation data sample through the multi-dimensional operation element mining model to generate a multi-dimensional operation element vector sample.

And 230, taking the fault trend test sample as a processing object of a trend expression element mining model, and performing element vector mining on the fault trend test sample through the trend expression element mining model to generate a trend expression element vector test sample.

Step 240, invoking a target fault diagnosis algorithm to determine a matching coefficient of the multi-dimensional operation element vector test sample and the trend expression element vector test sample, so as to obtain an operation fault matching viewpoint test sample, wherein the target fault diagnosis algorithm is used for expanding the multi-dimensional operation element vector test sample and the trend expression element vector test sample, and determining a matching coefficient of the multi-dimensional operation element vector test sample and the trend expression element vector test sample obtained by expansion.

And 250, determining a training error based on the operation fault matching viewpoint adjustment sample, and improving the variable of the target fault diagnosis algorithm through the training error.

In the fault diagnosis and optimization process of the crankshaft double-top vehicle control system, a series of training and debugging steps are usually adopted in the system in order to improve the accuracy and efficiency of a fault diagnosis algorithm. These steps include the acquisition, processing of operational data and fault trend samples, and the training and tuning of fault diagnosis algorithms using these samples.

Firstly, the system acquires an operation data test sample and a fault trend test sample. The operation data sample is a data sample generated when the system operates normally, and the fault trend sample reflects trend information of the operation data when faults occur. These examples are critical to understanding the normal behavior and failure modes of the system.

Secondly, the system takes the operation data sample as a processing object of the multi-dimensional operation element mining model. The multidimensional operation element mining model is a tool capable of deeply mining hidden information in operation data. The model can generate a multi-dimensional operation element vector sample by processing the operation data sample. These vectors numerically characterize the critical features of the operational data in various dimensions, which are an important basis for subsequent fault diagnosis.

Meanwhile, the system takes the fault trend adjustment sample as a processing object of the trend expression element mining model. The trend expression element mining model is specially used for extracting key information in fault trends. The model can generate a trend expression element vector test sample by processing the fault trend test sample. The vectors reflect key characteristics of fault trends in numerical values, and provide important basis for subsequent fault matching.

Then, the system calls a target fault diagnosis algorithm to determine the matching coefficients of the multidimensional operation element vector test sample and the trend expression element vector test sample. The target fault diagnosis algorithm is a tool specially used for matching the operation data with the fault trend. The algorithm can generate an operation fault matching viewpoint adjustment sample by determining the matching coefficient of the multi-dimensional operation element vector adjustment sample obtained by expansion and the trend expression element vector adjustment sample. The samples reflect the matching degree between the operation data and the fault trend in value, and are important indexes for evaluating the performance of the fault diagnosis algorithm.

Finally, the system determines a training error based on the operational fault matching perspective adjustment sample, and improves the variables of the target fault diagnosis algorithm through the training error. The training error reflects the performance of the fault diagnosis algorithm on the current sample. Through analysis and processing of the training errors, the system can find the defects in the algorithm and perform corresponding optimization adjustment. The training and optimizing process based on the sample can obviously improve the accuracy and efficiency of the fault diagnosis algorithm, so that the method can better meet the requirements in practical application.

In some examples, the multi-dimensional running element vector adjustment sample comprises P multi-dimensional running element vector adjustment samples, the trend indicative element vector adjustment sample comprises P trend indicative element vector adjustment samples, and P is an integer greater than 1. The step 240 of invoking the target fault diagnosis algorithm determines a matching coefficient of the multi-dimensional operation element vector adjustment sample and the trend expression element vector adjustment sample, to obtain an operation fault matching point of view adjustment sample, including: and calling a target fault diagnosis algorithm to determine matching coefficients of the P multidimensional operation element vector samples and the P trend expression element vector samples to obtain operation fault matching point set samples, wherein the operation fault matching point set samples comprise P times P matching coefficient samples, and the matching coefficient samples are used for reflecting the matching performance between the multidimensional operation element vector samples and the trend expression element vector samples.

In the fault diagnosis flow of the crankshaft double-top-car control system, when the matching of the multidimensional operation element vector test sample and the trend expression element vector test sample is involved, the system processes the test samples through a specific algorithm and generates a test sample reflecting the matching between the test samples.

First, the multidimensional operation element vector adjustment sample comprises P multidimensional operation element vector adjustment sample samples, and the samples represent key characteristics of the system in multiple dimensions under different operation states. Similarly, the trend expression element vector test sample also comprises P trend expression element vector test samples, and the test samples reflect key performances of the system under different fault trends.

Next, in step 240, the system invokes the target fault diagnosis algorithm to determine the matching coefficients for the P multidimensional operating element vector samples and the P trend indicative element vector samples. The purpose of this step is to quantify the degree of matching between each multidimensional run element vector sample and each trend indicative element vector sample. The matching coefficient is a numerical value reflecting the similarity or correlation of the two vector samples in terms of numerical characteristics.

In the process, the target fault diagnosis algorithm calculates each pair of multidimensional operation element vector test sample and trend expression element vector test sample to generate a matching coefficient test sample. Since there are P multidimensional running element vector samples and P trend performance element vector samples, a total of P matching coefficient samples are generated.

Finally, these matching coefficient tuning samples constitute an operational failure matching point of view set tuning sample. Each sample in the set reflects the matching between a multidimensional running element vector sample and a trend indicative element vector sample. Through the matching coefficient sample adjustment examples, the system can know the association degree between the multidimensional characteristics of the system and different fault trends under different running states, so that powerful data support is provided for subsequent fault diagnosis.

Fig. 2 shows a block diagram of an artificial intelligence fault diagnosis system 300, comprising: memory 310 for storing program instructions and data; a processor 320, coupled to the memory 310, executes instructions in the memory 310 to implement the methods described above.

Further, a computer storage medium is provided containing instructions which, when executed on a processor, implement the above-described method.

Therefore, aiming at the problems, the application provides a fault diagnosis method of the crankshaft double-top-vehicle control system based on multi-dimensional operation element mining and trend expression element mining. The method comprises the steps of comprehensively acquiring historical operation data (including sensor detection data, control output data and actuator state data) to be diagnosed, and utilizing a multidimensional operation element mining model to mine element vectors of the data to generate multidimensional operation element vectors. Meanwhile, the method also utilizes a trend expression element mining model to mine element vectors of a plurality of fault trend labels, and generates trend expression element vectors. By comparing the matching coefficients of the two types of vectors, the method can accurately identify the current fault trend of the system. Specifically, steps 110 through 150 of the present application constitute a complete fault diagnosis procedure. Step 110 ensures the comprehensiveness and accuracy of the data; step 120 and step 130 extract key features of the data by element vector mining; step 140, determining matching coefficients of the multidimensional operation element vector and the trend expression element vector through a target fault diagnosis algorithm; step 150 then determines a target fault trend label based on the matching coefficients. The series of steps are associated with each other and progressive layer by layer, so that the efficient and accurate fault diagnosis of the crankshaft double-top vehicle control system is realized.

More specifically, conventional methods and some existing data driving methods often rely on only a single sensor detection data and ignore control output data and actuator state data when dealing with fault diagnostics of a crankshaft dual overhead vehicle control system. A limitation of this approach is that it may not capture all of the characteristics of the fault, particularly when the fault involves interaction of multiple system components. Step 110 of the solution solves this problem, since it explicitly indicates that the data to be acquired includes not only sensor detection data but also control output data and actuator status data. Thus, the fault diagnosis process can consider more comprehensive system operation information, thereby improving the accuracy of diagnosis.

Existing fault diagnosis methods typically only consider fixed labels corresponding to known fault types when determining fault trends. This means that these methods may not be able to identify effectively when new, unknown fault types occur. Step 130 and step 150 in the technical solution provide a more flexible and extensible way to process fault trends by introducing a trend expression element mining model and a determination process of target fault trend labels. In particular, through element vector mining and matching coefficient calculation, the method can identify new fault trends different from the known labels, so that the adaptability and the universality of the method in practical application are enhanced.

Because of limitations of conventional methods and some existing data driven methods in terms of data processing and fault trend identification, their diagnostic accuracy may be compromised. Technical solution a more accurate and systematic fault diagnosis method is provided by a series of steps (element vector mining of step 120, matching coefficient determination of step 140 and target fault trend label determination of step 150). The method not only considers more comprehensive data, but also accurately identifies fault trend through element vector mining and matching coefficient calculation, thereby improving the accuracy of diagnosis.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The apparatus and method embodiments described above are merely illustrative, for example, flow diagrams and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, the functional modules in the embodiments of the present application may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a network device, or the like) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the same, but rather, various modifications and variations may be made by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims (10)

1. A fault diagnosis method for a crankshaft double-top vehicle control system, which is characterized by being applied to an artificial intelligent fault diagnosis system, the method comprising:

acquiring historical operation data to be diagnosed and X fault trend labels, wherein X is an integer not smaller than 1, and the historical operation data to be diagnosed comprises sensor detection data, control output data and actuator state data of a crankshaft double-top-vehicle control system;

taking the historical operation data to be diagnosed as a processing object of a multi-dimensional operation element mining model, and performing element vector mining on the historical operation data to be diagnosed through the multi-dimensional operation element mining model to generate multi-dimensional operation element vectors;

taking the X fault trend labels as processing objects of a trend expression element mining model, and performing element vector mining on the X fault trend labels through the trend expression element mining model to generate X trend expression element vectors;

Invoking a target fault diagnosis algorithm, determining matching coefficients of the multidimensional operation element vector and the X trend expression element vectors through the target fault diagnosis algorithm to obtain X operation fault matching views, wherein the target fault diagnosis algorithm is used for expanding the multidimensional operation element vector and the X trend expression element vectors and determining the matching coefficients of the target multidimensional operation element vector obtained by expansion and the X target trend expression element vectors;

and determining a target fault trend label from the X fault trend labels based on the X operation fault matching viewpoints, wherein the operation fault matching viewpoint corresponding to the target fault trend label is the maximum matching viewpoint weight in the X operation fault matching viewpoints.

2. The method for diagnosing faults of a double-top crankshaft vehicle control system according to claim 1, wherein the step of calling a target fault diagnosis algorithm, determining matching coefficients of the multidimensional operation element vector and the X trend expression element vectors through the target fault diagnosis algorithm to obtain X operation fault matching views includes:

determining a target knowledge relation network corresponding to the target fault diagnosis algorithm based on the target fault diagnosis algorithm;

Projecting the multi-dimensional operation element vector to the target knowledge relation network to obtain the target multi-dimensional operation element vector, wherein the characteristic size of the target multi-dimensional operation element vector is larger than that of the multi-dimensional operation element vector;

projecting the X trend expression element vectors to the target knowledge relation network to obtain X target trend expression element vectors, wherein the characteristic size of the target trend expression element vectors is larger than that of the trend expression element vectors;

and calling the target fault diagnosis algorithm to determine matching coefficients of the target multidimensional operation element vector and the X target trend expression element vectors, so as to obtain X operation fault matching views.

3. The fault diagnosis method for a crank double-top vehicle control system according to claim 2, wherein projecting the multidimensional operation element vector to the target knowledge relation network to obtain the target multidimensional operation element vector comprises:

obtaining a knowledge vector derivative model;

determining the characteristic size of the target knowledge relation network;

optimizing variables of the knowledge vector derivative model based on the characteristic size of the target knowledge relationship network to obtain a target knowledge vector derivative model corresponding to the target knowledge relationship network;

And taking the multidimensional operation element vector as a processing object of the target knowledge vector derivative model, and expanding the multidimensional operation element vector through the target knowledge vector derivative model to obtain the target multidimensional operation element vector, wherein the characteristic size of the target multidimensional operation element vector is consistent with the characteristic size of the target knowledge relationship network.

4. The method for diagnosing faults in a double-top crankshaft vehicle control system as claimed in claim 2, wherein said invoking the target fault diagnosis algorithm to perform a matching coefficient determination on the target multidimensional operating element vector and the X target trend representing element vectors to obtain X operating fault matching perspectives includes:

acquiring the characteristic size of the target multidimensional operation element vector and the characteristic size of each target trend expression element vector in the X target trend expression element vectors;

and carrying out operation fault decision on the target multidimensional operation element vector and each target trend expression element vector based on the characteristic size of the target multidimensional operation element vector and the characteristic size of each target trend expression element vector in the X target trend expression element vectors to obtain X operation fault matching views.

5. The fault diagnosis method for a crank double-overhead-vehicle control system according to claim 1, wherein the step of taking the historical operation data to be diagnosed as a processing object of a multi-dimensional operation element mining model, performing element vector mining on the historical operation data to be diagnosed through the multi-dimensional operation element mining model, generating a multi-dimensional operation element vector, comprises:

the historical operation data to be diagnosed are used as processing objects of a data disassembly sub-model in the multi-dimensional operation element mining model, and are disassembled through the data disassembly sub-model to obtain Y target operation data blocks, wherein Y is an integer greater than 1;

the Y target operation data blocks are used as processing objects of a distribution detection sub-model in the multi-dimensional operation element mining model, and the distribution characteristics of each target operation data block in the Y target operation data blocks are determined through the distribution detection sub-model, so that Y operation data block distribution characteristics are obtained;

and taking the Y target operation data blocks and the Y operation data block distribution characteristics as processing objects of a characteristic focusing sub-model in the multi-dimensional operation element mining model, and performing element extraction on the Y target operation data blocks and the Y operation data block distribution characteristics through the characteristic focusing sub-model to generate a multi-dimensional operation element vector.

6. The method for diagnosing a fault in a double-overhead-crankshaft-vehicle control system according to claim 5, wherein the generating a multidimensional operation element vector by using the Y target operation data blocks and the Y operation data block distribution features as processing objects of a feature focusing sub-model in the multidimensional operation element mining model and performing element extraction on the Y target operation data blocks and the Y operation data block distribution features by the feature focusing sub-model comprises:

the Y target operation data blocks are used as processing objects of feature element focusing nodes in the feature focusing sub-model, matching coefficients of every two target operation data blocks in the Y target operation data blocks are determined through the feature element focusing nodes, Z data block matching coefficients are obtained, wherein Z=Y (Y-1)/2, and Z is an integer larger than 1;

the Y target operation data blocks are used as processing objects of feature embedding nodes in the feature focusing sub-model, and element knowledge of each target operation data block in the Y target operation data blocks is extracted through the feature embedding nodes to obtain Y operation data block element knowledge;

Taking the Z data block matching coefficients, the Y operation data block element knowledge and the Y operation data block distribution characteristics as processing objects of feature integration nodes in the feature focusing sub-model, and carrying out element integration on the Z data block matching coefficients, the Y operation data block element knowledge and the Y operation data block distribution characteristics through the feature integration nodes to obtain a multidimensional operation element vector;

the step of taking the historical operation data to be diagnosed as a processing object of a data disassembly sub-model in the multi-dimensional operation element mining model, and disassembling the historical operation data to be diagnosed through the data disassembly sub-model to obtain Y target operation data blocks, wherein the step of obtaining the Y target operation data blocks comprises the following steps:

taking the historical operation data to be diagnosed as a processing object of a dismantling node in the data dismantling sub-model, and dismantling the historical operation data to be diagnosed through the dismantling node to generate Y operation data units;

taking the Y operation data units as processing objects of feature coding nodes in the data disassembly sub-model, and performing feature coding on the Y operation data units through the feature coding nodes to generate Y operation data block coding sets;

And taking the Y running data block code sets as processing objects of data updating nodes in the data disassembly sub-model, and carrying out data updating on the Y running data block code sets through the data updating nodes to generate Y target running data blocks, wherein the characteristic size corresponding to the target running data blocks is smaller than that of the running data units.

7. The method for diagnosing faults in a double-top-drive control system of claim 1, wherein the step of using the X fault trend labels as processing objects of a trend-expressing element mining model, and performing element vector mining on the X fault trend labels by the trend-expressing element mining model to generate X trend-expressing element vectors, comprises:

taking each fault trend label in the X fault trend labels as a processing object of a stage trend disassembly node in the trend expression element mining model, and carrying out stage trend disassembly on each fault trend label in the X fault trend labels through the stage trend disassembly node to obtain X stage trend sub-label sets;

the X stage trend sub-label sets are used as processing objects of trend vector identification sub-models in the trend expression element mining model, and each stage trend sub-label set in the X stage trend sub-label sets is processed through the trend vector identification sub-models to generate X stage trend characterization vector sets;

The X stage trend characterization vector sets are used as processing objects of stage trend concern sub-models in the trend expression element mining model, each stage trend characterization vector set in the X stage trend characterization vector sets is processed through the stage trend concern sub-models, and X stage trend concern vector sets are generated, wherein the stage trend concern vector sets comprise at least one stage trend concern vector;

integrating the phase trend attention vectors in each phase trend attention vector set in the X phase trend attention vector sets to obtain X fault trend integrated vectors;

performing interval numerical mapping on each fault trend integration vector in the X fault trend integration vectors to obtain X trend expression element vectors;

the step of using the X stage trend characterization vector sets as a processing object of a stage trend attention sub-model in the trend expression element mining model, and processing each stage trend characterization vector set in the X stage trend characterization vector sets through the stage trend attention sub-model to generate X stage trend attention vector sets includes:

Each stage trend representation vector set in the X stage trend representation vector sets is used as a processing object of a characteristic focusing node in the stage trend focusing sub-model, semantics among the stage trend representation vectors in each stage trend representation vector set are identified through the characteristic focusing node, and X first fault representation focusing vector sets are obtained, wherein the first fault representation focusing vector sets comprise at least one first fault representation focusing vector;

taking each first failure expression focusing vector set in the X first failure expression focusing vector sets as a processing object of a first fusion node in the stage trend focusing sub-model, and carrying out integrated processing on the first failure expression focusing vector in each first failure expression focusing vector set through the first fusion node to obtain X first fusion vectors;

taking the X first fusion vectors as a processing object of a first interval value mapping node in the stage trend concern submodel, and carrying out interval value mapping on the X first fusion vectors through the first interval value mapping node to obtain X first quantized trend vectors;

And taking the X first quantized trend vectors as processing objects of feature embedding nodes in the stage trend attention sub-model, and identifying semantic information between every two first quantized trend vectors in the X first quantized trend vectors through the feature embedding nodes to obtain X stage trend attention vector sets, wherein each stage trend attention vector set comprises X (X-1)/2 stage trend attention vectors.

8. The method for diagnosing a failure of a crankshaft dual overhead vehicle control system according to claim 1, further comprising:

acquiring an operation data sample adjustment sample and a fault trend sample adjustment sample, wherein the fault trend sample adjustment sample is used for reflecting fault trend information of the operation data sample adjustment sample;

taking the operation data sample as a processing object of a multi-dimensional operation element mining model, and performing element vector mining on the operation data sample through the multi-dimensional operation element mining model to generate a multi-dimensional operation element vector sample;

taking the fault trend test sample as a processing object of a trend expression element mining model, and performing element vector mining on the fault trend test sample by the trend expression element mining model to generate a trend expression element vector test sample;

Invoking a target fault diagnosis algorithm to determine matching coefficients of the multi-dimensional operation element vector sample and the trend expression element vector sample to obtain an operation fault matching viewpoint sample, wherein the target fault diagnosis algorithm is used for expanding the multi-dimensional operation element vector sample and the trend expression element vector sample and determining matching coefficients of the multi-dimensional operation element vector sample and the trend expression element vector sample obtained by expansion;

determining a training error based on the operation fault matching viewpoint adjustment sample, and improving the variable of the target fault diagnosis algorithm through the training error;

the multi-dimensional operation element vector test sample comprises P multi-dimensional operation element vector test sample cases, the trend expression element vector test sample cases comprise P trend expression element vector test sample cases, and P is an integer greater than 1; the target fault diagnosis algorithm is invoked to determine a matching coefficient of the multidimensional operation element vector sample and the trend expression element vector sample, so as to obtain an operation fault matching viewpoint sample, which comprises the following steps: and calling a target fault diagnosis algorithm to determine matching coefficients of the P multidimensional operation element vector samples and the P trend expression element vector samples to obtain operation fault matching point set samples, wherein the operation fault matching point set samples comprise P times P matching coefficient samples, and the matching coefficient samples are used for reflecting the matching performance between the multidimensional operation element vector samples and the trend expression element vector samples.

9. An artificial intelligence fault diagnosis system, comprising: a memory for storing program instructions and data; a processor coupled to a memory for executing instructions in the memory to implement the method of any of claims 1-8.

10. A computer storage medium containing instructions which, when executed on a processor, implement the method of any of claims 1-8.