Intelligent system for automatically judging automobile condition

Through natural language processing and neural network model combined with particle swarm algorithm and network monitoring tools, an automated judgment system is built, which solves the problems of inefficient and prone to errors in vehicle condition judgment, and achieves fast and accurate judgment of vehicle condition.

CN119004097AActive Publication Date: 2024-11-22BEIJING JINGZHENG INFORMATION TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202410977383.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-22
Estimated Expiration
2044-07-22

AI Technical Summary

Technical Problem

In the prior art, vehicle condition judgments rely on manual inspections, are inefficient and prone to errors, lack intelligent and automated analysis systems, and it is difficult to process massive historical text data.

Method used

Using natural language processing (NLP) technology and neural network model, combined with particle swarm algorithm and network monitoring tools, an automated judgment system is built, vehicle data is obtained through data acquisition sensors, text conversion and preprocessing is performed, and convolutional neural networks and cyclic nerves are used. The network model makes vehicle condition judgment.

Benefits of technology

It realizes fast and accurate automated judgment of vehicle conditions, improves judgment efficiency and accuracy, and reduces the human error rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004097A_ABST
    Figure CN119004097A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent system for automatically judging the vehicle condition, and the system comprises a collection module which is used for collecting the vehicle condition data of a vehicle and carrying out the text conversion, and obtaining the text data; and the first judgment module is used for inputting the text data into a pre-trained vehicle condition judgment model and outputting a vehicle condition judgment result. Automatic analysis and judgment of vehicle text data are realized by using a natural language processing (NLP) technology and a neural network model, and the problems of low efficiency and error proneness in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle management, and in particular to an intelligent system for automatically judging the condition of a vehicle. Background Art

[0002] At present, the judgment of vehicle condition mainly relies on manual inspection and analysis, which is inefficient and prone to human errors. With the increase in the number of vehicles and the accumulation of vehicle condition data, how to use intelligent technology to quickly and accurately determine the vehicle condition has become an urgent problem to be solved. In the existing technology, there is a lack of a system that can effectively analyze and learn massive historical text data, resulting in a low level of automation and intelligence in vehicle condition judgment. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems in the above-mentioned technology to a certain extent. To this end, the purpose of the present invention is to propose an intelligent system for automatically judging the vehicle condition, which uses natural language processing (NLP) technology and neural network models to realize automatic analysis and judgment of vehicle text data, and solve the problems of low efficiency and easy error in the prior art.

[0004] To achieve the above objectives, an embodiment of the present invention provides an intelligent system for automatically determining the vehicle condition of a vehicle, comprising:

[0005] The acquisition module is used to collect the vehicle condition data of the vehicle and perform text conversion to obtain text data;

[0006] The first judgment module is used to input the text data into a pre-trained vehicle condition judgment model and output a vehicle condition judgment result.

[0007] According to some embodiments of the present invention, the acquisition module includes:

[0008] The first building block is used to:

[0009] Determine the interval information between various data collection sensors that collect vehicle condition data; the interval information includes distance information obtained by GPS positioning, communication delay and network hop number;

[0010] A minimum distance matrix is ​​constructed based on the interval information and the particle swarm algorithm. The minimum distance matrix is ​​a square matrix, in which each data acquisition node represents the minimum acquisition cost between two corresponding data acquisition sensors, as a data acquisition network;

[0011] The first determination module is used to:

[0012] Get the total bandwidth corresponding to the data collection network;

[0013] Determine a first number of receiving paths and a second number of forwarding paths for each data collection node;

[0014] According to the first quantity and the second quantity, a preset allocation data table is searched to determine the important coefficient of each data collection node in the minimum distance matrix;

[0015] The allocation module is used to allocate the total bandwidth according to the important coefficient of each data collection node to obtain the allocation result; the data collection network collects the vehicle condition data of the vehicle according to the allocation result;

[0016] The conversion module is used to perform text conversion on the vehicle data to obtain text data.

[0017] According to some embodiments of the present invention, the various data acquisition sensors include an air flow sensor, an ABS sensor, a throttle position sensor, a crankshaft position sensor, a camshaft position sensor, a water temperature sensor, an oil pressure sensor, an oxygen sensor, a vehicle speed sensor and a knock sensor.

[0018] According to some embodiments of the present invention, the first determination module obtains the total bandwidth corresponding to the data collection network in real time based on Nagi os, Zabbix, and PRTG Network Monitor network monitoring tools.

[0019] According to some embodiments of the present invention, the allocation module allocates the total bandwidth according to the importance coefficient of each data acquisition node to obtain an allocation result, including:

[0020] Determine the proportion information according to the important coefficient of each data collection node;

[0021] The total bandwidth is allocated according to the ratio information to obtain an allocation result.

[0022] According to some embodiments of the present invention, the first judgment module obtains a vehicle condition judgment model, including:

[0023] An acquisition module is used to acquire the historical vehicle condition data of the vehicle and perform text conversion to obtain the historical text data of the vehicle;

[0024] The data preprocessing module is used to clean and format the historical text data, remove noise and irrelevant information, standardize the data format, and obtain preprocessed data;

[0025] The NLP processing module is used to perform word segmentation on the preprocessed data using a professional word segmentation dictionary, break the preprocessed data into independent phrases, classify the phrases, and obtain classified data;

[0026] A neural network training module is used to train a combined model of a convolutional neural network and a recurrent neural network based on classified data to obtain a training model;

[0027] The evaluation module is used to evaluate the training model based on the test data, and when the evaluation is qualified, a vehicle condition judgment model is obtained.

[0028] According to some embodiments of the present invention, the NLP processing module includes:

[0029] The second building module is used to obtain vehicle component information and build a professional word segmentation dictionary for vehicle components according to the vehicle component information;

[0030] The word segmentation module is used to perform word segmentation on the preprocessed data based on a professional word segmentation dictionary, breaking the preprocessed data into independent phrases;

[0031] The classification module is used to perform part-of-speech tagging on the phrases, and classify the phrases according to the tagging results to obtain classification data.

[0032] According to some embodiments of the present invention, the data preprocessing module includes:

[0033] A cleaning module is used to obtain type information of historical text data, query a preset data cleaning rule file according to the type information, obtain corresponding data cleaning rules, and perform data cleaning according to the corresponding data cleaning rules to obtain cleaned data;

[0034] The formatting processing module is used to process the cleaned data according to the adapted regular expression to obtain pre-processed data.

[0035] According to some embodiments of the present invention, the classification module includes:

[0036] The second determination module is used to determine the word vector of each phrase, match the word vector with the standard word vector, and perform part-of-speech tagging on the phrase according to the part-of-speech corresponding to the matched standard word vector;

[0037] The sorting module is used to classify and sort the phrases of the same part of speech to obtain classified data.

[0038] According to some embodiments of the present invention, the evaluation module comprises:

[0039] The calculation module is used to obtain the H test text data included in the test data and the sample vehicle condition determination results corresponding to the H test text data, input the H test text data into the training model respectively to output H corresponding predicted vehicle condition determination results, and calculate the matching degree Q between the sample vehicle condition determination results corresponding to the H test text data and the H predicted vehicle condition determination results:

[0040]

[0041] Among them, S a ' is the ath predicted vehicle condition determination result output by the training model; S a The sample vehicle condition determination result of the a-th test text data input into the training model;

[0042] The second judgment module is used to judge whether the matching degree is greater than a preset matching degree. When it is determined that the matching degree is greater than the preset matching degree, it means that the evaluation is qualified and a vehicle condition judgment model is obtained; otherwise, it means that the evaluation is unqualified and the training model needs to be further trained.

[0043] The present invention proposes an intelligent system for automatically judging the condition of a car, which uses natural language processing (NLP) technology and a neural network model to realize automatic analysis and judgment of vehicle text data, thereby solving the problems of low efficiency and easy errors in the prior art.

[0044] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0045] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 is a block diagram of an intelligent system for automatically determining a vehicle condition according to an embodiment of the present invention;

[0048] Figure 2 is a block diagram of a collection module according to an embodiment of the present invention;

[0049] Figure 3 is a block diagram of a first judgment module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0051] like Figure 1 As shown, the embodiment of the present invention provides an intelligent system for automatically judging the vehicle condition of a vehicle, including:

[0052] The acquisition module is used to collect the vehicle condition data of the vehicle and perform text conversion to obtain text data;

[0053] The first judgment module is used to input the text data into a pre-trained vehicle condition judgment model and output a vehicle condition judgment result.

[0054] The working principle of the above technical solution is: based on the acquisition module, the vehicle condition data of the car is collected and converted into text to obtain text data; based on the first judgment module, the text data is input into a pre-trained vehicle condition judgment model to output the vehicle condition judgment result.

[0055] The beneficial effects of the above technical solution are: using natural language processing (NLP) technology and neural network models to realize automatic analysis and judgment of vehicle text data, solving the problems of low efficiency and easy errors in the existing technology.

[0056] like Figure 2 As shown, according to some embodiments of the present invention, the acquisition module includes:

[0057] The first building block is used to:

[0058] Determine the interval information between various data collection sensors that collect vehicle condition data; the interval information includes distance information obtained by GPS positioning, communication delay and network hop number;

[0059] A minimum distance matrix is ​​constructed based on the interval information and the particle swarm algorithm. The minimum distance matrix is ​​a square matrix, in which each data acquisition node represents the minimum acquisition cost between two corresponding data acquisition sensors, as a data acquisition network;

[0060] The first determination module is used to:

[0061] Get the total bandwidth corresponding to the data collection network;

[0062] Determine a first number of receiving paths and a second number of forwarding paths for each data collection node;

[0063] According to the first quantity and the second quantity, a preset allocation data table is searched to determine the important coefficient of each data collection node in the minimum distance matrix;

[0064] The allocation module is used to allocate the total bandwidth according to the important coefficient of each data collection node to obtain the allocation result; the data collection network collects the vehicle condition data of the vehicle according to the allocation result;

[0065] The conversion module is used to perform text conversion on the vehicle data to obtain text data.

[0066] Working principle of the above technical solution: In this embodiment, the distance information, communication delay and network hop number between each data sensor are determined based on the interval information; in order to reduce the communication delay, high-speed, low-latency communication protocols and technologies can be used. At the same time, optimizing the network layout and reducing network congestion are also effective ways to reduce communication delay. In the vehicle network, the number of network hops and the delay of data transmission can be reduced by multi-hop communication.

[0067] In this embodiment, data acquisition node: each data acquisition sensor is regarded as a node. Collection cost: varies under the influence of interval information. Particle swarm algorithm: PSO is an optimization algorithm that finds the optimal solution to the problem by simulating the behavior of bird flocks foraging. Each particle is regarded as a possible network configuration, its position represents the connection mode between nodes, and the speed represents the adjustment direction of the connection mode. Steps to build the minimum distance matrix: Step 1: Initialization: Node list: List all data acquisition sensors (nodes). Interval matrix: Create a matrix in which the element M[i][j] represents the physical distance between node i and node j (considering the weighting of communication delay). Particle swarm: Initialize a group of particles, each particle represents a possible network configuration (i.e., the connection mode between nodes). Step 2: Evaluate fitness and calculate the total cost: For each particle (i.e., each network configuration), calculate its total cost. This usually involves traversing the interval matrix and selecting appropriate paths and connections based on the configuration of the particles, and then accumulating the costs on these paths. Fitness function: The total cost is used as the output of the fitness function. The goal is to find the configuration with the lowest total cost. Step 3: Update particles: Personal best position: For each particle, update its personal best position (i.e., the best configuration found so far). Global best position: Find the global best position among all particles (i.e., the best configuration found in the entire particle swarm). Speed ​​and position update: Update the speed and position of each particle according to the rules of PSO. This usually involves moving a certain distance to the personal best position and the global best position. Step 4: Iteration: Repeat steps 2 and 3 until a certain stopping condition is met (such as reaching the maximum number of iterations, the change in fitness value is less than a certain threshold, etc.). Step 5: Construct the minimum distance matrix: After the PSO algorithm converges, a globally optimal network configuration is obtained. Based on this configuration, construct the minimum distance matrix. The "minimum distance matrix" here is actually a matrix that represents the minimum cost between nodes in the optimal network configuration (considering physical distance and communication delay). Each element D[i][j] in the matrix can represent the cost of the minimum cost path from node i to node j (in the optimal network configuration).

[0068] In this embodiment, the total bandwidth corresponding to the data acquisition network is obtained based on the first determination module; the receiving path is the path for receiving data; and the forwarding path is the path for forwarding data. The first number of receiving paths and the second number of forwarding paths of each data acquisition node are determined; the allocation data table is a data table for querying important coefficients according to the first number and the second number. The preset allocation data table is queried according to the first number and the second number to determine the important coefficient of each data acquisition node in the minimum distance matrix; that is, the importance of each data acquisition node is determined. The allocation module is used to allocate the total bandwidth according to the important coefficient of each data acquisition node to obtain an allocation result; the data acquisition network collects the vehicle condition data of the car according to the allocation result; the conversion module is used to convert the vehicle data into text to obtain text data.

[0069] The beneficial effects of the above technical solution are: constructing a minimum distance matrix based on interval information and particle swarm algorithm, constructing an optimized data acquisition network, and allocating corresponding bandwidth to each data acquisition node in the data acquisition network, further optimizing the network, and facilitating improving the speed and accuracy of data acquisition.

[0070] According to some embodiments of the present invention, the various data acquisition sensors include an air flow sensor, an ABS sensor, a throttle position sensor, a crankshaft position sensor, a camshaft position sensor, a water temperature sensor, an oil pressure sensor, an oxygen sensor, a vehicle speed sensor and a knock sensor.

[0071] Air flow sensor: used to measure the amount of air inhaled by the engine and convert the information into an electrical signal and input it into the electronic control unit (ECU) as a reference signal for injection timing.

[0072] ABS sensor: used to monitor vehicle speed. In case of emergency braking, the wheel speed is fed back to the brake system, which then controls the restrained rotation of the wheel to achieve the best braking effect.

[0073] Throttle position sensor: used to detect the opening of the throttle valve. It is linked to the throttle valve through a lever mechanism, measures the angle voltage signal of the throttle valve opening, and transmits the signal to the ECU as a reference signal for correction of the injection amount and ignition advance angle.

[0074] Crankshaft position sensor: used to detect the top dead center signal, crankshaft angle signal and engine speed signal, and provide them to the ECU as a reference signal to determine the ignition timing, injection timing and working order.

[0075] Camshaft position sensor: provides crankshaft angle reference position signal as the main control signal for injection timing control and ignition timing control.

[0076] Water temperature sensor: installed on the cylinder block, used to detect the temperature information of the engine cooling water, and convert the information into electrical signals and provide it to the ECU.

[0077] Oil pressure sensor: used to detect the oil pressure and send out an alarm signal when the pressure is insufficient.

[0078] Oxygen sensor: installed on the exhaust pipe, used to detect the oxygen content in the engine exhaust, determine whether the gasoline and air are completely burned, and perform closed-loop control of the injection amount.

[0079] Vehicle speed sensor: detects the vehicle's speed and provides a vehicle speed signal to the ECU for cruise control and speed limit and fuel cut-off control. It is also the main control signal of the automatic transmission.

[0080] Knock sensor: detects the vibration of the engine cylinder to allow the electronic controller to identify the engine knock condition. When the engine knocks, the engine vibration is transmitted to the crystal through the mass in the sensor. The piezoelectric crystal generates voltage on the two poles due to the pressure generated by the vibration of the mass, converting the vibration into a voltage signal output.

[0081] In addition, there are intake temperature sensors, fuel level sensors, coolant level sensors, battery level sensors, fuel rail fuel pressure sensors, fuel tank pressure sensors, transmission oil pressure sensors, collision sensors, ultrasonic ranging sensors, corner sonar (corner radar) sensors, brake pressure switch sensors, brake light switch sensors, distance sensors, intake manifold absolute pressure sensors, boost pressure sensors, cylinder combustion pressure sensors, tire pressure detection sensors, wheel speed sensors, deceleration sensors, electric seat sensors, headlight high and low beam control sensors and other sensors. They each undertake different vehicle condition data collection tasks and jointly ensure the safe, stable and efficient operation of the vehicle.

[0082] According to some embodiments of the present invention, the first determination module obtains the total bandwidth corresponding to the data collection network in real time based on Nagios, Zabbix, and PRTG NetworkMonitor network monitoring tools.

[0083] Working principle and beneficial effects of the above technical solutions: Nagios is an open source network and system monitoring tool that can monitor network traffic, server performance, application status, etc. It uses a plug-in system to extend its functions, including monitoring network bandwidth. Zabbix is ​​a powerful enterprise-level open source monitoring solution for monitoring various IT components, including network bandwidth. It can monitor network traffic, bandwidth utilization and other indicators in real time, and provide rich chart and report functions. PRTG Network Monitor is a graphical network monitoring software that supports multiple functions such as bandwidth monitoring and traffic analysis. It provides an intuitive graphical interface to help administrators quickly understand the usage of network bandwidth. Installation and configuration: Install Nagios, Zabbix or PRTG Network Monitor on key nodes of the data collection network. Configure the monitoring tool to connect to the network interface to be monitored. Set monitoring indicators, including total bandwidth, bandwidth utilization, etc. Real-time monitoring: Use the real-time monitoring function of the monitoring tool to continuously collect network bandwidth data. The monitoring tool will automatically calculate and display the current total bandwidth usage. Based on the network monitoring tool, the total bandwidth corresponding to the data collection network is obtained in real time to improve the accuracy of the obtained data.

[0084] According to some embodiments of the present invention, the allocation module allocates the total bandwidth according to the importance coefficient of each data acquisition node to obtain an allocation result, including:

[0085] Determine the proportion information according to the important coefficient of each data collection node;

[0086] The total bandwidth is allocated according to the ratio information to obtain an allocation result.

[0087] The working principle and beneficial effects of the above technical solution are as follows: determining the proportion information according to the important coefficient of each data acquisition node; allocating the total bandwidth according to the proportion information to obtain the allocation result, so as to allocate accurate bandwidth to each data acquisition node and improve the data acquisition rate.

[0088] like Figure 3 As shown, according to some embodiments of the present invention, the first judgment module obtains a vehicle condition judgment model, including:

[0089] An acquisition module is used to acquire the historical vehicle condition data of the vehicle and perform text conversion to obtain the historical text data of the vehicle;

[0090] The data preprocessing module is used to clean and format the historical text data, remove noise and irrelevant information, standardize the data format, and obtain preprocessed data;

[0091] The NLP processing module is used to perform word segmentation on the preprocessed data using a professional word segmentation dictionary, break the preprocessed data into independent phrases, classify the phrases, and obtain classified data;

[0092] A neural network training module is used to train a combined model of a convolutional neural network and a recurrent neural network based on classified data to obtain a training model;

[0093] The evaluation module is used to evaluate the training model based on the test data, and when the evaluation is qualified, a vehicle condition judgment model is obtained.

[0094] The working principle and beneficial effects of the above technical solution: The acquisition module is responsible for obtaining the historical vehicle condition data of the vehicle from various data sources (such as the vehicle manufacturer's database, the vehicle OBD interface, the third-party data provider, etc.), and converting the data into text format for subsequent processing. Use API interface or data crawling technology to obtain data from the data source. The data may exist in binary, XML, JSON or other formats and need to be converted to plain text format. During the conversion process, it may be necessary to parse and organize the data to ensure that all necessary information is correctly extracted and formatted. The data preprocessing module cleans the historical text data, removes noise (such as irrelevant characters, repeated information, etc.), and standardizes the data format for subsequent processing. Convert data types to ensure data consistency and accuracy. The NLP processing module uses a professional word segmentation dictionary to perform word segmentation on the preprocessed text data, decomposes the text into independent phrases, and categorizes these phrases. Use natural language processing (NLP) libraries (such as Python's NLTK, jieba, etc.) for word segmentation. Customize or use existing word segmentation dictionaries to ensure that vehicle-related professional terms can be accurately identified. Use part-of-speech tagging, named entity recognition and other technologies to further analyze phrases. Design classification algorithms or rules to classify phrases into different categories (such as engine condition, tire condition, fuel consumption, etc.). The neural network training module uses a combined model of convolutional neural network (CNN) and recurrent neural network (RNN) based on the classified data to extract features and learn the relationship between vehicle conditions and text data. Design a hybrid model that combines CNN and RNN. CNN is used to capture local features of text (such as the relationship between phrases), and RNN (especially LSTM or GRU) is used to process long-term dependencies in sequence data. Divide the classified data into training sets and validation sets. Use deep learning frameworks (such as TensorFlow, PyTorch) to build and train the model. Adjust model parameters such as the number of layers, number of neurons, learning rate, etc. to optimize model performance. The evaluation module uses test data to evaluate the trained model to verify its accuracy and reliability. If the evaluation result is qualified, a vehicle condition judgment model that can be used for actual vehicle condition judgment is obtained. It is easy to obtain an accurate vehicle condition judgment model.

[0095] According to some embodiments of the present invention, the NLP processing module includes:

[0096] The second building module is used to obtain vehicle component information and build a professional word segmentation dictionary for vehicle components according to the vehicle component information;

[0097] The word segmentation module is used to perform word segmentation on the preprocessed data based on a professional word segmentation dictionary, breaking the preprocessed data into independent phrases;

[0098] The classification module is used to perform part-of-speech tagging on the phrases, and classify the phrases according to the tagging results to obtain classification data.

[0099] Working principle and beneficial effects of the above technical solution: The second construction module is responsible for obtaining vehicle component information and constructing a professional word segmentation dictionary for vehicle components based on this information. This dictionary will be used for subsequent word segmentation processing to ensure accurate identification of vehicle-related professional terms. Implementation method: Data collection: Collect vehicle component information from vehicle manufacturers, maintenance manuals, technical documents or professional databases. Dictionary construction: Construct a professional word segmentation dictionary based on the sorted component information. Each entry in the dictionary should be a professional term related to vehicle components. Dictionary update: With the development of new models and new technologies, the word segmentation dictionary is regularly updated to ensure its accuracy and comprehensiveness. The word segmentation module performs word segmentation processing on the preprocessed text data based on the professional word segmentation dictionary, and breaks the text into independent phrases. Implementation method: Loading dictionary: Before word segmentation processing, load the constructed professional word segmentation dictionary. Word segmentation algorithm: Use word segmentation algorithms (such as maximum positive matching, minimum segmentation, etc.) in combination with professional word segmentation dictionaries to segment the text. The algorithm should give priority to the entries in the dictionary to ensure accurate identification of professional terms. Result output: Output the word segmentation results as an independent phrase list for subsequent processing. The classification module performs part-of-speech tagging on the phrases output by the word segmentation module, and classifies the phrases according to the tagging results to obtain classified data. Implementation method: Part-of-speech tagging: Use part-of-speech tagging tools (such as HanLP, jieba, etc.) to tag the phrases after word segmentation. The tagging results will include the part of speech of each phrase (such as noun, verb, adjective, etc.). Category classification: Classify the phrases according to the part-of-speech tagging results and preset category rules. For example, phrases in the noun category can be classified as vehicle parts, fault types, etc.; phrases in the verb category can be classified as operating behaviors, fault manifestations, etc. Rule formulation: The rules for category classification can be formulated according to actual needs, including but not limited to keyword-based matching, part-of-speech-based screening, context-based understanding, etc. Result output: Output the classified data into a structured format, such as JSON, XML or a database table, for subsequent processing and analysis.

[0100] According to some embodiments of the present invention, the data preprocessing module includes:

[0101] A cleaning module is used to obtain type information of historical text data, query a preset data cleaning rule file according to the type information, obtain corresponding data cleaning rules, and perform data cleaning according to the corresponding data cleaning rules to obtain cleaned data;

[0102] The formatting processing module is used to process the cleaned data according to the adapted regular expression to obtain pre-processed data.

[0103] The working principle and beneficial effects of the above technical solution: The cleaning module obtains the type information of historical text data (such as log type, report type, etc.), and then queries the preset data cleaning rule file based on this type information to obtain the corresponding data cleaning rules, and applies these rules to clean the data to remove noise, irrelevant information and erroneous data to obtain the cleaned data. Implementation method: Type information acquisition: Extract type information from the metadata, file name, file path or other identifiers of the text data. Rule file query: According to the acquired type information, find the corresponding cleaning rules in the preset data cleaning rule file. These rule files can be in XML, JSON, YAML and other formats, which contain cleaning logic and parameters for different types of data. Rule application: Load the found cleaning rules and use them to process the original text data. The processing process may include but is not limited to: Removing useless characters, spaces and line breaks. Replacing erroneous data or inconsistently formatted data. Deleting or correcting incomplete records. Identifying and deleting duplicate data. Result output: Save the cleaned data as a new file or data structure for use by subsequent modules. The formatting processing module uses an adapted regular expression to further process the cleaned data to ensure that the format of the data meets the requirements of subsequent analysis or model training. Formatting processing may include standardization of data format, extraction and reorganization of fields, etc. Implementation method: Regular expression design: Design an adapted regular expression based on the characteristics of the data and the needs of subsequent processing. These expressions are used to match and extract specific information in the text, such as date, time, and numerical values. Data processing: Use regular expressions to match and extract the cleaned data to separate the required information from the text. According to needs, the extracted information is formatted or standardized, such as converting date strings to a unified date format, converting numerical strings to floating point numbers or integers, etc. Reorganize the data and reorganize the extracted information into a new data structure according to a predetermined format, such as a JSON object, a database record, etc. Result output: The formatted data is output as preprocessed data for use by subsequent modules such as the NLP processing module and the neural network training module. Through the collaborative work of these two modules, the system can effectively convert raw historical text data into clean, uniformly formatted preprocessed data suitable for subsequent processing, laying a solid foundation for subsequent data analysis and model training.

[0104] According to some embodiments of the present invention, the classification module includes:

[0105] The second determination module is used to determine the word vector of each phrase, match the word vector with the standard word vector, and perform part-of-speech tagging on the phrase according to the part-of-speech corresponding to the matched standard word vector;

[0106] The sorting module is used to classify and sort the phrases of the same part of speech to obtain classified data.

[0107] The working principle and beneficial effects of the above technical solution: The second determination module determines the word vector for each phrase, and matches these word vectors with the word vectors in the standard word vector library. According to the matching results, the part of speech corresponding to the matched standard word vector is used to tag the phrase with the part of speech. Implementation method: Word vector determination: Use word embedding technology (such as Word2Vec, GloVe, BERT, etc.) to determine a word vector for each phrase obtained after word segmentation. The word vector is a numerical representation of a phrase in a vector space, which can capture the semantic relationship between phrases. Word vector matching: The calculated word vector is similar to the word vector in the pre-trained standard word vector library (such as cosine similarity). The standard word vector library contains a large number of phrases with known parts of speech and their corresponding word vectors. Part of speech tagging: According to the similarity calculation results, find the standard word vector that is most similar to the phrase to be tagged, and obtain the part of speech corresponding to the standard word vector. Then, the part of speech is used as the part of speech tagging result of the phrase to be tagged. The sorting module is responsible for categorizing and sorting the phrases with the part of speech tagged by the second determination module. According to the part of speech of the phrase and other possible context information, phrases with the same or similar categories are classified together to form classified data. Implementation method: Category definition: The classification category system needs to be defined. These categories can be divided based on actual business scenarios such as vehicle parts, fault types, and operating behaviors. Classification rule formulation: According to the part of speech of the phrase and possible context information (such as the part of speech of adjacent phrases, semantic relationships, etc.), formulate the rules for phrase classification. These rules can be rule-based classification algorithms (such as decision trees, rule engines, etc.) or statistical classification algorithms (such as naive Bayes, support vector machines, etc.). Phrase classification: Apply the formulated classification rules to classify the phrases marked with part of speech into corresponding categories. Result output: The classified data is sorted into a structured format (such as JSON, XML, database table, etc.) and output as part of the preprocessed data for use by subsequent modules. Through the collaborative work of the second determination module and the sorting module, the system can realize the part of speech tagging and category classification of the phrases after word segmentation, providing a more accurate and useful data basis for subsequent data analysis and model training.

[0108] According to some embodiments of the present invention, the evaluation module comprises:

[0109] The calculation module is used to obtain the H test text data included in the test data and the sample vehicle condition determination results corresponding to the H test text data, input the H test text data into the training model respectively to output H corresponding predicted vehicle condition determination results, and calculate the matching degree Q between the sample vehicle condition determination results corresponding to the H test text data and the H predicted vehicle condition determination results:

[0110]

[0111] Among them, S a ' is the ath predicted vehicle condition determination result output by the training model; S a The sample vehicle condition determination result of the a-th test text data input into the training model;

[0112] The second judgment module is used to judge whether the matching degree is greater than a preset matching degree. When it is determined that the matching degree is greater than the preset matching degree, it means that the evaluation is qualified and a vehicle condition judgment model is obtained; otherwise, it means that the evaluation is unqualified and the training model needs to be further trained.

[0113] The working principle and beneficial effects of the above technical solution are as follows: based on the calculation module, the test data includes H test text data and the sample vehicle condition determination results corresponding to the H test text data, the H test text data are respectively input into the training model to output H corresponding predicted vehicle condition determination results, the matching degree between the sample vehicle condition determination results corresponding to the H test text data and the H predicted vehicle condition determination results is calculated, and it is judged whether the matching degree is greater than the preset matching degree. When it is determined that the matching degree is greater than the preset matching degree, it indicates that the evaluation is qualified and the vehicle condition determination model is obtained; otherwise, it indicates that the evaluation is unqualified and the training model needs to be further trained. Accurately judge whether the trained model is qualified.

[0114] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent system for automatically judging the condition of a car, characterized in that: include: The acquisition module is used to collect the vehicle condition data of the vehicle and perform text conversion to obtain text data; The first judgment module is used to input the text data into a pre-trained vehicle condition judgment model and output a vehicle condition judgment result.

2. The intelligent system for automatically determining the vehicle condition of a vehicle as claimed in claim 1, characterized in that: The acquisition module comprises: The first building block is used to: Determine the interval information between various data collection sensors that collect vehicle condition data; the interval information includes distance information obtained by GPS positioning, communication delay and network hop number; A minimum distance matrix is ​​constructed based on the interval information and the particle swarm algorithm. The minimum distance matrix is ​​a square matrix, in which each data acquisition node represents the minimum acquisition cost between two corresponding data acquisition sensors, as a data acquisition network; The first determination module is used to: Get the total bandwidth corresponding to the data collection network; Determine a first number of receiving paths and a second number of forwarding paths for each data collection node; According to the first quantity and the second quantity, a preset allocation data table is searched to determine the important coefficient of each data collection node in the minimum distance matrix; The allocation module is used to allocate the total bandwidth according to the important coefficient of each data collection node to obtain the allocation result; the data collection network collects the vehicle condition data of the vehicle according to the allocation result; The conversion module is used to perform text conversion on the vehicle data to obtain text data.

3. The intelligent system for automatically determining the vehicle condition of a vehicle as claimed in claim 2, characterized in that: The various data acquisition sensors include air flow sensor, ABS sensor, throttle position sensor, crankshaft position sensor, camshaft position sensor, water temperature sensor, oil pressure sensor, oxygen sensor, vehicle speed sensor and knock sensor.

4. The intelligent system for automatically determining the vehicle condition as claimed in claim 2, characterized in that: The first determination module obtains the total bandwidth corresponding to the data collection network in real time based on Nagios, Zabbix, and PRTG Network Monitor network monitoring tools.

5. The intelligent system for automatically determining the vehicle condition of a vehicle as claimed in claim 2, characterized in that: The allocation module allocates the total bandwidth according to the important coefficient of each data acquisition node to obtain an allocation result, including: Determine the proportion information according to the important coefficient of each data collection node; The total bandwidth is allocated according to the ratio information to obtain an allocation result.

6. The intelligent system for automatically determining the vehicle condition of claim 1, characterized in that: The first judgment module obtains a vehicle condition judgment model, including: An acquisition module is used to acquire the historical vehicle condition data of the vehicle and perform text conversion to obtain the historical text data of the vehicle; The data preprocessing module is used to clean and format the historical text data, remove noise and irrelevant information, standardize the data format, and obtain preprocessed data; The NLP processing module is used to perform word segmentation on the preprocessed data using a professional word segmentation dictionary, break the preprocessed data into independent phrases, classify the phrases, and obtain classified data; A neural network training module is used to train a combined model of a convolutional neural network and a recurrent neural network based on classified data to obtain a training model; The evaluation module is used to evaluate the training model based on the test data, and when the evaluation is qualified, a vehicle condition judgment model is obtained.

7. The intelligent system for automatically determining the vehicle condition of claim 6, characterized in that: The NLP processing module comprises: The second building module is used to obtain vehicle component information and build a professional word segmentation dictionary for vehicle components according to the vehicle component information; The word segmentation module is used to perform word segmentation on the preprocessed data based on a professional word segmentation dictionary, breaking the preprocessed data into independent phrases; The classification module is used to perform part-of-speech tagging on the phrases, and classify the phrases according to the tagging results to obtain classification data.

8. The intelligent system for automatically determining the vehicle condition of claim 6, characterized in that: Data preprocessing module, including: A cleaning module is used to obtain type information of historical text data, query a preset data cleaning rule file according to the type information, obtain corresponding data cleaning rules, and perform data cleaning according to the corresponding data cleaning rules to obtain cleaned data; The formatting processing module is used to process the cleaned data according to the adapted regular expression to obtain pre-processed data.

9. The intelligent system for automatically determining the vehicle condition of claim 7, characterized in that: The classification module comprises: The second determination module is used to determine the word vector of each phrase, match the word vector with the standard word vector, and perform part-of-speech tagging on the phrase according to the part-of-speech corresponding to the matched standard word vector; The sorting module is used to classify and sort the phrases of the same part of speech to obtain classified data.

10. The intelligent system for automatically determining the vehicle condition of claim 6, characterized in that: The evaluation module comprises: The calculation module is used to obtain the H test text data included in the test data and the sample vehicle condition determination results corresponding to the H test text data, input the H test text data into the training model respectively to output H corresponding predicted vehicle condition determination results, and calculate the matching degree Q between the sample vehicle condition determination results corresponding to the H test text data and the H predicted vehicle condition determination results: Among them, S a ' is the ath predicted vehicle condition determination result output by the training model; S a The sample vehicle condition determination result of the a-th test text data input into the training model; The second judgment module is used to judge whether the matching degree is greater than a preset matching degree. When it is determined that the matching degree is greater than the preset matching degree, it means that the evaluation is qualified and a vehicle condition judgment model is obtained; otherwise, it means that the evaluation is unqualified and the training model needs to be further trained.

Citation Information

Patent Citations

  • Intelligent algorithm system for vehicle condition analysis

    CN112232056A

  • NLP analysis identification and data cleaning method based on artificial intelligence

    CN114626346A

  • Engineering optimization solving method based on Gaussian distribution QPSO algorithm

    CN115688854A

  • Vehicle-mounted fault analysis method and device, computer equipment and storage medium

    CN116737882A

  • Applied artificial intelligence for natural language processing automotive reporting system

    US20200020179A1