Vehicle full life cycle management method and system based on OBD data

Abstract

The invention belongs to the technical field of data management, and discloses a vehicle full life cycle management method and system based on OBD data. The method comprises the following steps: a cloud data center deploys a block chain network, defines a vehicle full life cycle, and collects historical OBD data; the cloud data center performs clustering processing and key feature screening; the cloud data center constructs a data classification model, a data analysis model and a management strategy generation model; the cloud data center receives the real-time OBD data and performs data classification by using a data classification model; the cloud data center is used for carrying out data analysis on the real-time OBD data; the cloud data center is used for generating a management strategy by using a management strategy generation model; and the cloud data center performs distributed storage by using a block chain network. According to the invention, the problems of data islands, insufficient data security, insufficient data analysis capability and low intelligent degree in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data management, and particularly relates to a vehicle full-life cycle management method and system based on OBD data. Background Art

[0002] With the rapid development of the automotive industry, vehicle management has become an important topic. Vehicle full-life cycle management covers the entire process from vehicle design, manufacturing, sales, use, maintenance to scrapping. In this process, the On-Board Diagnostics (OBD) system plays a crucial role. The OBD system monitors the data of key components such as the engine, catalytic converter, and oxygen sensor in real time, and is a system that monitors the vehicle status and controls emissions, ensuring the normal operation of the vehicle emission system. The OBD data generated can reflect the true state of the vehicle. Therefore, how to implement vehicle full-life cycle management based on OBD data has become an important development direction in the automotive industry.

[0003] In the field of vehicle full-life cycle management, although the OBD system provides a large amount of OBD data, there are still many deficiencies in the existing technology for effectively managing OBD data:

[0004] 1) There is a problem of data islands: OBD data is often stored in isolation in their respective vehicles or repair stations, lacking a unified data platform for integration and analysis. There is a problem of data islands, making it difficult to achieve comprehensive and unified management, which leads to a waste of data resources and low utilization efficiency;

[0005] 2) Insufficient data security: OBD data involves vehicle operating status and owner privacy. The existing technology lacks effective encryption and privacy protection measures during data storage, and is insufficient to cope with potential data leakage and abuse risks, and is vulnerable to data leakage and malicious attacks;

[0006] 3) Insufficient data analysis capabilities: Existing OBD data analysis methods are often based on simple statistics and analysis, making it difficult to mine the deep value and potential rules in the data, and having limited processing capabilities for OBD data, making it difficult to meet the rapid processing and analysis requirements of large-scale and real-time data, resulting in lagging management decisions;

[0007] 4) Low degree of intelligence: Existing data management methods mainly rely on manual experience and regular inspections, lacking the generation of intelligent management strategies based on artificial intelligence, and it is difficult to achieve precise and efficient management. Summary of the Invention

[0008] To solve the problems of data silos, insufficient data security, insufficient data analysis capabilities, and low intelligence level existing in the prior art, the purpose of the present invention is to provide a vehicle full-life cycle management method and system based on OBD data.

[0009] The technical solution adopted by the present invention is as follows:

[0010] A vehicle full-life cycle management method based on OBD data, comprising the following steps:

[0011] The cloud data center deploys a blockchain network, defines several life cycle stages of the vehicle full-life cycle, and collects historical OBD data of several life cycle stages;

[0012] The cloud data center performs clustering processing and key feature screening on several historical OBD data to obtain historical OBD data clusters of several life cycle stages and corresponding key feature indicators;

[0013] The cloud data center constructs a data classification model, a data analysis model, and a management strategy generation model using artificial intelligence algorithms according to several historical OBD data clusters and corresponding key feature indicators;

[0014] The cloud data center receives real-time OBD data sent by the OBD system, and uses the data classification model to classify the real-time OBD data to obtain the corresponding target life cycle stage;

[0015] The cloud data center performs data analysis on the real-time OBD data based on the target key feature indicators of the target life cycle stage using the data analysis model to obtain real-time data analysis results;

[0016] The cloud data center generates a management strategy using the management strategy generation model according to the real-time data analysis results to obtain a real-time management strategy;

[0017] The cloud data center distributes and stores the real-time OBD data, the corresponding target life cycle stage, and the real-time data analysis results using the blockchain network according to the real-time management strategy.

[0018] Further, the cloud data center deploys a blockchain network, defines several life cycle stages of the vehicle full-life cycle, and collects historical OBD data of several life cycle stages, including the following steps:

[0019] Using blockchain technology, all data servers of the cloud data center are distributedly connected as data nodes to obtain a blockchain network;

[0020] According to the identity allocation mechanism, a number of data nodes in the blockchain network are divided into a number of consensus nodes and a number of storage nodes to obtain a data storage sub-network and a blockchain consensus sub-network;

[0021] Define a number of life cycle stages of the vehicle full life cycle; the life cycle stages include the purchase stage, the running-in stage, the stable stage, the aging stage, and the scrapping stage;

[0022] Collect a number of historical OBD data for each life cycle stage, and preprocess the number of historical OBD data to obtain a number of preprocessed historical OBD data belonging to different life cycle stages.

[0023] Furthermore, the cloud data center performs clustering processing and key feature screening on a number of historical OBD data to obtain historical OBD data clusters and corresponding key feature indicators for a number of life cycle stages, including the following steps:

[0024] The cloud data center uses the AP clustering algorithm to cluster a number of preprocessed historical OBD data belonging to different life cycle stages to obtain a number of cluster centers;

[0025] Set the corresponding life cycle stage for each cluster center, and obtain the Euclidean distance between each preprocessed historical OBD data and the cluster centers of all life cycle stages;

[0026] Divide all preprocessed historical OBD data into the cluster center with the closest Euclidean distance to obtain historical OBD data clusters for different life cycle stages;

[0027] Use the RF algorithm to obtain the importance score of each feature index of the preprocessed historical OBD data corresponding to the cluster center, and use the feature indexes with the top several importance scores as key feature indexes;

[0028] Traverse the preprocessed historical OBD data corresponding to all cluster centers to obtain the key feature indicators of historical OBD data clusters for a number of life cycle stages.

[0029] Furthermore, the cloud data center uses an artificial intelligence algorithm to construct a data classification model, a data analysis model, and a management strategy generation model according to a number of historical OBD data clusters and corresponding key feature indicators, including the following steps:

[0030] Construct a data classification model using a deep learning algorithm according to a number of historical OBD data clusters and corresponding life cycle stages;

[0031] Set the corresponding attention weight value for each historical OBD data cluster according to the key feature indicators of the historical OBD data cluster;

[0032] Based on a number of historical OBD data clusters and corresponding attention weight values, use deep learning algorithms to construct a data analysis model, and generate a number of historical weighted fusion features and corresponding historical data analysis results;

[0033] Based on a number of historical weighted fusion features and corresponding historical data analysis results, use reinforcement learning algorithms to construct a management strategy generation model, and generate a number of historical management strategy generation experiences.

[0034] Furthermore, the data classification model is constructed based on the LSTM-Elman algorithm, and the data classification model includes a semantic data feature extraction module constructed based on the LSTM algorithm and a data classification module constructed based on the Elman algorithm that are connected in sequence;

[0035] The data analysis model is constructed based on the DBN-Attention-MLP algorithm, and the data analysis model includes a deep feature extraction module constructed based on the DBN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the MLP algorithm that are connected in sequence;

[0036] The management strategy generation model is constructed based on the MPO-MOGRPO algorithm, and the management strategy generation model includes a meta-policy optimization module constructed based on the MPO algorithm and a management strategy generation module constructed based on the MOGRPO algorithm. The management strategy generation module includes a set of objective functions, an experience replay pool, a policy network, and an agent. The agent is respectively connected to the set of objective functions, the experience replay pool, and the policy network. The meta-policy optimization module is respectively connected to the policy network and the experience replay pool.

[0037] Furthermore, based on a number of historical OBD data clusters and corresponding life cycle stages, use deep learning algorithms to construct a data classification model, including the following steps:

[0038] Use the life cycle stage as the true label of a number of preprocessed historical OBD data in the corresponding historical OBD data cluster;

[0039] Divide a number of preprocessed historical OBD data with true labels in all historical OBD data clusters into a training sample set and a test sample set according to a ratio of 7:3;

[0040] Use the LSTM-Elman algorithm to construct an initial data classification model, and input the training sample set to optimize and train the initial data classification model to obtain an optimized data classification model;

[0041] Input the test sample set to optimize and train the initial data classification model to obtain a number of predicted labels of the optimized data classification model;

[0042] Compare and statistically analyze several predicted labels and corresponding true labels to obtain the test accuracy of the optimized data classification model;

[0043] If the test accuracy is greater than the accuracy threshold, output the final data classification model; otherwise, continue with the optimization training of the data classification model.

[0044] Furthermore, based on several historical OBD data clusters and corresponding attention weight values, use deep learning algorithms to construct a data analysis model and generate several historical weighted fusion features and corresponding historical data analysis results, including the following steps:

[0045] Use the DBN-Attention-MLP algorithm to construct an initial data analysis model, and update the initial data analysis model according to the attention weight values of the historical OBD data clusters to obtain an updated data analysis model;

[0046] Input the corresponding historical OBD data clusters, optimize and train the updated data analysis model to obtain an optimized data analysis model, and generate several historical weighted fusion features and corresponding historical data analysis results of the current historical OBD data clusters;

[0047] Traverse all historical OBD data clusters, optimize and train the optimized data analysis model to obtain a final data analysis model, and generate several historical weighted fusion features and corresponding historical data analysis results of all historical OBD data clusters.

[0048] Furthermore, based on several historical weighted fusion features and corresponding historical data analysis results, use reinforcement learning algorithms to construct a management strategy generation model and generate several historical management strategy generation experiences, including the following steps:

[0049] Use the MPO-MOGRPO algorithm to construct an initial management strategy generation model; the initial management strategy generation model includes an initial meta-policy optimization module and an initial management strategy generation module;

[0050] Take the life cycle stage as the scenario for meta-policy optimization, and train the initial meta-policy optimization module according to several historical weighted fusion features in different scenarios to obtain a final meta-policy optimization module;

[0051] Use the final meta-policy optimization module to update the initial policy network of the initial management strategy generation module in different scenarios to obtain an updated policy network;

[0052] Set a set of target functions and an experience replay pool for the initial management strategy generation module with the updated policy network, and set an action space and a state space for the agent of the initial management strategy generation module;

[0053] Take the management strategy generation problem as a simulation environment, and obtain an optimized management strategy generation module according to the updated policy network and the agent with an action space and a state space set up;

[0054] Traverse all the objective functions in the set of objective functions, and optimize and train the optimized management strategy generation module according to the analysis results of several historical data to obtain the final management strategy generation module, and generate several historical management strategy generation experiences;

[0055] Integrate the final meta-strategy optimization module and the final management strategy generation module to obtain the final management strategy generation model, and store several historical management strategy generation experiences in the experience replay pool.

[0056] Furthermore, the cloud data center, according to the real-time management strategy, uses the blockchain network to perform distributed storage on the real-time OBD data, the corresponding target life cycle stage, and the real-time data analysis results, including the following steps:

[0057] The cloud data center, according to the real-time management strategy, performs data sharding on the real-time OBD data to obtain several real-time data shards including replica shards;

[0058] According to the real-time management strategy, use a random encryption algorithm to encrypt several real-time data shards to obtain several encrypted real-time data shards, and generate a real-time data storage request;

[0059] Use the blockchain consensus sub-network of the blockchain network to perform consensus on the real-time data storage request. If the consensus is successful, proceed to the next step;

[0060] According to the real-time management strategy, use the data storage sub-network of the blockchain network to perform distributed storage on several encrypted real-time data shards, and return several real-time storage addresses;

[0061] Synchronize several real-time data shards, the corresponding real-time storage addresses, the target life cycle stage, and the real-time data analysis results to the distributed ledger.

[0062] A vehicle full life cycle management system based on OBD data is used to implement the vehicle full life cycle management method. The system is set in the cloud data center, and the system includes an initialization unit, a data processing unit, a model construction unit, a data classification unit, a data analysis unit, a management strategy generation unit, and a distributed storage unit. The cloud data center is respectively communicatively connected to the OBD systems of several vehicles.

[0063] The beneficial effects of the present invention are:

[0064] The present invention discloses a vehicle full - life - cycle management method and system based on OBD data, which defines several stages of the vehicle full - life - cycle, and collects historical OBD data, conducts data analysis, and formulates strategies for different stages, making the management strategies more targeted and effective, and improving the data management effect; a cloud data center is established to realize the centralized storage, management, and analysis of OBD data. Through the data sharing and exchange mechanism, data islands are broken, and cross - vehicle, cross - brand, and cross - regional data integration is achieved, improving the data utilization efficiency; a blockchain network constructed by blockchain technology is used to encrypt and store OBD data to ensure the security and immutability of the data; data classification models and data analysis models are constructed based on artificial intelligence algorithms to achieve more accurate and comprehensive data analysis, improving the analysis ability of OBD data, ensuring the real - time nature of data management, and adjusting model parameters based on different life - cycle stages to improve the accuracy of data analysis; a management - strategy generation model is constructed based on artificial intelligence algorithms to dynamically generate and optimize management strategies, making the management strategies more flexible and adaptable, improving the intelligence level of data management, avoiding reliance on manual experience and regular inspections, and achieving precise and efficient management.

[0065] Other beneficial effects of the present invention will be further described in the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is a flowchart of the vehicle full - life - cycle management method based on OBD data in the present invention.

[0067] Figure 2 is a structural block diagram of the vehicle full - life - cycle management system based on OBD data in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0069] Embodiment 1:

[0070] As Figure 1 shown, this embodiment provides a vehicle full - life - cycle management method based on OBD data, including the following steps:

[0071] S1: In the cloud data center, deploy a blockchain network, define several life - cycle stages of the vehicle full - life - cycle, and collect historical OBD data of several life - cycle stages, including the following steps:

[0072] S1 - 1: Use blockchain technology to distributively connect all data servers in the cloud data center as data nodes to obtain a blockchain network;

[0073] S1-2: According to the identity allocation mechanism, divide several data nodes of the blockchain network into several consensus nodes and several storage nodes to obtain a data storage sub-network and a blockchain consensus sub-network;

[0074] The blockchain consensus sub-network is used to use the Practical Byzantine Fault Tolerance (PBFT) consensus algorithm to consensus on data storage requests; the data storage sub-network is used to perform distributed storage of data shards;

[0075] S1-3: Define several life cycle stages of the vehicle full life cycle; the life cycle stages include the purchase stage, the running-in stage, the stable stage, the aging stage, and the scrapping stage;

[0076] In the purchase stage, focus on vehicle basic information, initial state data, etc.; in the running-in stage, focus on engine operation data, special maintenance records during the running-in period, etc.; in the stable stage, focus on vehicle operation data, and focus on fault warnings and regular maintenance records; in the aging stage, focus on the storage and analysis of data such as vehicle component wear and performance degradation; in the scrapping stage, focus on the final state data before vehicle scrapping, the scrapping reason, etc.;

[0077] S1-4: Collect several historical OBD data of each life cycle stage, and preprocess the several historical OBD data to obtain several preprocessed historical OBD data belonging to different life cycle stages;

[0078] The preprocessing includes formatting to ensure that the data meets the input requirements of subsequent model training, and data cleaning to remove redundant and irrelevant information, improving the quality and training efficiency of the data;

[0079] S2: The cloud data center performs clustering processing and key feature screening on several historical OBD data to obtain historical OBD data clusters of several life cycle stages and corresponding key feature indicators, including the following steps:

[0080] S2-1: The cloud data center uses the Affinity-Propagation (AP) clustering algorithm to cluster several preprocessed historical OBD data belonging to different life cycle stages to obtain several cluster centers, including the following steps:

[0081] S2-1-1: Initialize the AP clustering algorithm to obtain N initial cluster centers; N is the total number of cluster centers;

[0082] S2-1-2: Introduce a reward and punishment mechanism to obtain the reputation value of each preprocessed historical OBD data in several preprocessed historical OBD data belonging to different life cycle stages;

[0083] The formula for the credibility value is as follows:

[0084]

[0085] In the formula, C i" is the current credibility value of each data; i" is the data indicator; is the initial credibility value of each data; α1 is the weight of the reward mechanism; R(i") is the single - time reward value of the data; α2 is the weight of the punishment mechanism; P(i") is the single - time punishment value of the data; N is the total number of data;

[0086] The formula for the reward and punishment mechanism is as follows:

[0087]

[0088] In the formula, ε is the number of times the data is selected as the clustering center; e is the natural constant;

[0089]

[0090] In the formula, bad i" is the number of abnormal data; B is the threshold of the number of times the punishment is triggered;

[0091] S2 - 1 - 3: According to the credibility value, perform a descending - power sorting on all pre - processed historical OBD data, and set the optimal bias parameter for the first N pre - processed historical OBD data;

[0092] S2 - 1 - 4: Introduce an iterative decay coefficient to update the attraction information and attribution information of all pre - processed historical OBD data, and obtain the updated attraction information and updated attribution information;

[0093] The formula for the attraction information is as follows:

[0094]

[0095] In the formula, r t'+1 (i", k") is the degree to which data k" is suitable as the clustering center of data i" at iteration t'+1, that is, the attraction information of data k" to data i"; a t； (i", j") is the degree of suitability for data i" to select data j" as its clustering center at iteration t'; r t' (i", j") is the degree to which data j" is suitable as the clustering center of data i" at iteration t'; s(i", k") is the similarity of data k" as the clustering center of data i"; i", j", and k" are all data indicators;

[0096] The formula for the attribution information is as follows:

[0097]

[0098] where r t'+1 (k", k") is the degree of suitability of the overall data k" as a clustering center at the iteration number t'+1; ∑ j"≠i",k" max{r t'+1 (j", k"), 0} is the degree of suitability of the data k" as a clustering center other than the data i" at the iteration number t'+1; a t'+1 (i", k") is the degree of suitability for the data i" to select the data k" as its clustering center at the iteration number t'+1, that is, the membership information of the data k" to the data i";

[0099] The iterative update formulas for the attractiveness information and the membership information are as follows:

[0100] r' t'+1 (i", k") = λ' * r t' (i", k") + (1 - λ') * r t'+1 (i", k")

[0101] a' t'+1 (i", k") = λ' * a t (i", k") + (1 - λ') * a t'+1 (i", k")

[0102] where r' t'+1 (i", k"), a' t'+1 (i", k") are the updated attractiveness information and the updated membership information of the data k" to the data i" at the iteration number t'+1; λ' is the iterative attenuation coefficient; r t' (i", k"), r t'+1 (i", k") are the attractiveness information of the data k" to the data i" at the iteration numbers t' and t'+1; a t' (i", k"), a t'+1 (i", k") are the membership information of the data k" to the data i" at the iteration numbers t' and t'+1;

[0103] S2-1-5: Update the N initial clustering centers according to the updated attractiveness information and the updated membership information to obtain N updated clustering centers;

[0104] The judgment formula for the clustering center is:

[0105] k" = argmax{a(i", k") + r(i", k")}

[0106] Wherein, both i" and k" are data indication quantities; if i" = k", then data i" is the clustering center of data k"; if i" ≠ k", then data k" is not the clustering center of data i".

[0107] S2-2: Set corresponding life cycle stages for each clustering center, and obtain the Euclidean distances between each preprocessed historical OBD data and the clustering centers of all life cycle stages.

[0108] S2-3: Divide all preprocessed historical OBD data into the clustering center with the closest Euclidean distance to obtain historical OBD data clusters of different life cycle stages.

[0109] S2-4: Use the Random Forest (RF) algorithm to obtain the importance scores of each feature index of the preprocessed historical OBD data corresponding to the clustering center, and use the feature indexes with the top several importance scores as key feature indexes, including the following steps:

[0110] S2-4-1: Use the RF algorithm to obtain the Gini coefficient of each feature index of the preprocessed historical OBD data corresponding to the clustering center, and use the Gini coefficient as the corresponding feature contribution degree.

[0111] The formula is:

[0112]

[0113] Wherein, is the feature contribution degree of the j"th feature; is the feature contribution degree of the j"th feature in the i"th decision tree of the random forest; i" is the decision tree indication quantity; j" is the feature indication quantity; n" is the total number of decision trees;

[0114]

[0115] Wherein, is the change amount of the Gini index before and after the branching of the m"th decision tree node; m" is the decision tree node indication quantity; M is the total number of decision tree nodes;

[0116]

[0117] Wherein, GI m" 、GI l 、GI r are the Gini indexes of the m"th, lth, and rth decision tree nodes; p m"k is the proportion of the kth category in the m"th decision tree node; k is the category indication quantity; |K| is the total number of categories;

[0118] S2-4-2: Normalize the feature contribution degree of each feature to obtain the normalized feature contribution degree, and based on the normalized feature contribution degree, obtain the feature selection standard value of each feature index of the preprocessed historical OBD data corresponding to the clustering center;

[0119] The formula is:

[0120]

[0121] In the formula, VIM' j" is the normalized feature contribution degree of the j-th feature; J is the total number of features; is the feature contribution degree of the j-th feature;

[0122]

[0123] In the formula, CFC j is the feature selection standard value of the j-th feature; VIM j* is the normalized feature contribution degree of the j*-th feature; j* is the feature indicator;

[0124] S2-4-3: Use the feature selection standard value of each feature index of the preprocessed historical OBD data as the importance score of the key features, and use the top R feature indexes with the importance scores as the key feature indexes;

[0125] S2-5: Traverse the preprocessed historical OBD data corresponding to all clustering centers to obtain the key feature indexes of the historical OBD data clusters in several life cycle stages;

[0126] S3: The cloud data center constructs a data classification model, a data analysis model, and a management strategy generation model using an artificial intelligence algorithm based on several historical OBD data clusters and the corresponding key feature indexes, including the following steps:

[0127] S3-1: Construct a data classification model using a deep learning algorithm based on several historical OBD data clusters and the corresponding life cycle stages;

[0128] The data classification model is constructed based on the Long Short-Term Memory (LSTM)-Elman algorithm, and the data classification model includes a semantic data feature extraction module constructed based on the LSTM algorithm and a data classification module constructed based on the Elman algorithm connected in sequence;

[0129] The LSTM in the semantic data feature extraction module is a special recurrent neural network, specifically designed to process and predict long-term dependencies in time series data. In vehicle full-life cycle management, OBD data is often time series data, and LSTM can effectively capture the temporal features in this data. The semantic features extracted by LSTM are richer and more accurate, which helps to improve the accuracy of subsequent data classification. The Elman in the data classification module is a simple recurrent neural network that adds feedback connections to the feedforward neural network, enabling the network to process dynamically changing data. The Elman network is responsible for classifying the data after LSTM feature extraction to determine the life cycle stage to which the data belongs. Due to the feedback connection of the Elman network, it can adapt to the dynamic changes of the input data, thereby improving the response speed and accuracy of the classification model for real-time OBD data.

[0130] According to a number of historical OBD data clusters and their corresponding life cycle stages, use deep learning algorithms to construct a data classification model, including the following steps:

[0131] S3-1-1: Use the life cycle stage as the true label of several preprocessed historical OBD data in the corresponding historical OBD data cluster.

[0132] S3-1-2: Divide the several preprocessed historical OBD data with true labels in all historical OBD data clusters into a training sample set and a test sample set according to a ratio of 7:3.

[0133] S3-1-3: Use the LSTM-Elman algorithm to construct an initial data classification model, input the training sample set, and optimize and train the initial data classification model to obtain an optimized data classification model.

[0134] S3-1-4: Input the test sample set, optimize and train the initial data classification model to obtain several predicted labels of the optimized data classification model.

[0135] S3-1-5: Compare and statistically analyze several predicted labels and their corresponding true labels to obtain the test accuracy of the optimized data classification model.

[0136] S3-1-6: If the test accuracy is greater than the accuracy threshold, output the final data classification model; otherwise, continue to optimize and train the data classification model.

[0137] S3-2: Set corresponding attention weight values for each historical OBD data cluster according to the key feature indicators of the historical OBD data clusters.

[0138] S3-3: Based on a number of historical OBD data clusters and their corresponding attention weight values, use deep learning algorithms to construct a data analysis model, and generate a number of historical weighted fusion features and corresponding historical data analysis results;

[0139] The data analysis model is constructed based on the Deep Belief Network (DBN)-Attention-Multilayer Perceptron (MLP) algorithm, and the data analysis model includes a deep feature extraction module constructed based on the DBN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the MLP algorithm, which are connected in sequence;

[0140] The deep feature extraction module uses the multi-layer structure of the DBN to perform layer-by-layer feature learning on the OBD data, extracts deep and abstract feature representations from the original data, and can capture complex structures and patterns in the data, thus providing more powerful and accurate feature representations; The Attention mechanism can assign different attention weights to different parts according to the importance or relevance of the input data, and is used to set which parts of the OBD data are most critical to the analysis task. By assigning higher weights to key features, the Attention mechanism helps the model focus more on important features, thereby improving the analysis accuracy; The MLP of the data analysis module is a feedforward neural network that can perform non-linear classification or regression analysis on the input data, and is responsible for the final analysis and prediction of the weighted fusion features after being processed by the DBN and Attention;

[0141] Based on a number of historical OBD data clusters and their corresponding attention weight values, use deep learning algorithms to construct a data analysis model, and generate a number of historical weighted fusion features and corresponding historical data analysis results, including the following steps:

[0142] S3-3-1: Use the DBN-Attention-MLP algorithm to construct an initial data analysis model, and update the initial data analysis model according to the attention weight values of the historical OBD data clusters to obtain an updated data analysis model;

[0143] S3-3-2: Input the corresponding historical OBD data clusters, optimize and train the updated data analysis model to obtain an optimized data analysis model, and generate a number of historical weighted fusion features and corresponding historical data analysis results of the current historical OBD data clusters;

[0144] S3-3-3: Traverse all historical OBD data clusters, optimize and train the optimized data analysis model to obtain the final data analysis model, and generate several historical weighted fusion features of all historical OBD data clusters and the corresponding historical data analysis results;

[0145] S3-4: According to several historical weighted fusion features and the corresponding historical data analysis results, use the reinforcement learning algorithm to construct a management policy generation model, and generate several historical management policy generation experiences;

[0146] The management policy generation model is constructed based on the Meta-Policy Optimization (MPO)-Deep Q Network (DQN)-Multi-Objective Group Relative Policy Optimization (MOGRPO) algorithm. The management policy generation model includes a meta-policy optimization module constructed based on the MPO algorithm and a management policy generation module constructed based on the MOGRPO algorithm. The management policy generation module includes a set of objective functions, an experience replay pool, a policy network, and an agent. The agent is respectively connected to the set of objective functions, the experience replay pool, and the policy network. The meta-policy optimization module is respectively connected to the policy network and the experience replay pool;

[0147] The meta-policy optimization module is used to optimize the network parameters of the policy network in the management policy generation module so that these parameters can quickly adapt to the weighted fusion features of new and unseen OBD data, improving the generalization ability of the model. Even under unseen OBD data, the policy network can be updated based on previous learning experiences, improving the adaptability of the management policy generation model. The set of objective functions of the management policy generation module can handle multiple conflicting objectives, such as storage resource utilization rate, storage resource scheduling efficiency, storage encryption level, etc., and generate management policies that balance these objectives. The agent learns historical management policies through the experience replay pool, continuously optimizing its own policy generation ability. The agent controls the policy network according to the learned experiences to generate more effective management policies. The design of the experience replay pool and the agent enables the model to continuously learn and optimize, improving the quality of policy generation. Since the management policy generation module adopts a group exploration method, it can avoid falling into local optimal solutions to a certain extent. The policy network outputs the distribution probability of actions in a given state. The management policy generation module directly updates the policy network through gradients, eliminating the Critic model in traditional reinforcement learning, making the algorithm structure more concise;

[0148] According to a number of historical weighted fusion features and the corresponding historical data analysis results, use the reinforcement learning algorithm to construct a management strategy generation model and generate a number of historical management strategy generation experiences, including the following steps:

[0149] S3-4-1: Use the MPO-MOGRPO algorithm to construct an initial management strategy generation model; the initial management strategy generation model includes an initial meta-policy optimization module and an initial management strategy generation module;

[0150] S3-4-2: Take the life cycle stage as the scenario for meta-policy optimization, and train the initial meta-policy optimization module according to a number of historical weighted fusion features in different scenarios to obtain the final meta-policy optimization module;

[0151] S3-4-3: Use the final meta-policy optimization module to update the initial policy network of the initial management strategy generation module in different scenarios to obtain an updated policy network;

[0152] S3-4-4: Set a set of target functions and an experience replay pool for the initial management strategy generation module with the updated policy network, and set an action space and a state space for the agent of the initial management strategy generation module;

[0153] S3-4-5: Take the management strategy generation problem as a simulation environment, and obtain an optimized management strategy generation module according to the updated policy network and the agent with the action space and state space set;

[0154] S3-4-6: Traverse all the target functions in the set of target functions, and optimize and train the optimized management strategy generation module according to a number of historical data analysis results to obtain the final management strategy generation module and generate a number of historical management strategy generation experiences;

[0155] S3-4-7: Integrate the final meta-policy optimization module and the final management strategy generation module to obtain the final management strategy generation model, and store a number of historical management strategy generation experiences in the experience replay pool;

[0156] S4: The cloud data center receives the real-time OBD data sent by the OBD system, and uses the data classification model to classify the real-time OBD data to obtain the corresponding target life cycle stage, including the following steps:

[0157] S4-1: The cloud data center receives the real-time OBD data sent by the OBD system and preprocesses the real-time OBD data to obtain preprocessed real-time OBD data;

[0158] S4-2: Use the semantic data feature extraction module of the data classification model to extract the real-time semantic features of the preprocessed real-time OBD data;

[0159] S4-3: Use the data classification module of the data classification model to classify the data according to the real-time semantic features and obtain the corresponding target life cycle stage;

[0160] S5: The cloud data center uses the data analysis model based on the target key feature indicators of the target life cycle stage to perform data analysis on the real-time OBD data and obtain the real-time data analysis result, including the following steps:

[0161] S5-1: The cloud data center extracts the corresponding target key feature indicators according to the target life cycle stage of the real-time OBD data. These target key feature indicators reflect the important data features in the target life cycle stage; for example, the relevant features of data such as vehicle component wear and performance degradation in the aging stage, including features such as engine operating status, vehicle speed, and fuel consumption;

[0162] S5-2: Generate the real-time attention weight value of the target life cycle stage according to the target key feature indicators;

[0163] S5-3: Use the deep feature extraction module of the data analysis model to extract the real-time deep features of the preprocessed real-time OBD data;

[0164] S5-4: According to the real-time attention weight value, use the attention weight module of the data analysis model to perform weighted fusion on the real-time deep features to obtain the real-time weighted fusion features;

[0165] S5-5: According to the real-time weighted fusion features, use the data analysis module of the data analysis model to perform data analysis and obtain the real-time data analysis result;

[0166] The real-time data analysis result has different focuses according to different life cycle stages. For example, the real-time data analysis result in the aging stage includes the vehicle component wear prediction result and the performance degradation prediction result;

[0167] S6: The cloud data center uses the management strategy generation model according to the real-time data analysis result to generate the management strategy and obtain the real-time management strategy, including the following steps:

[0168] S6-1: The cloud data center uses the meta-strategy optimization module of the management strategy generation model to update the strategy network of the management strategy generation module according to the real-time weighted fusion features corresponding to the real-time data analysis result, and obtains the updated strategy network;

[0169] S6-2: Analyze the real-time data analysis results to obtain several real-time data analysis states, and based on the several real-time data analysis states, update the state space of the agents in the management policy generation module of the management policy generation model to obtain an updated state space;

[0170] S6-3: Randomly extract several historical management policy generation experiences from the experience replay pool of the management policy generation module, and based on the several historical management policy generation experiences, generate several possible management decision actions;

[0171] S6-4: According to the several possible management decision actions, update the action space of the agents in the management policy generation module to obtain an updated action space;

[0172] S6-5: Select a real-time objective function from the set of objective functions of the management policy generation module, and based on the real-time objective function, use the agent to control the updated policy network to generate the probability distribution of all possible management decision actions in the updated action space corresponding to each real-time data analysis state in the updated state space;

[0173] S6-6: Take the possible management decision action with the highest probability distribution in the updated action space as the execution management decision action corresponding to the real-time data analysis state;

[0174] S6-7: Integrate the execution management decision actions of all real-time data analysis states in the updated state space to obtain a real-time management policy;

[0175] The real-time management policy includes real-time replica sharding decision, real-time random encryption decision, and real-time storage resource scheduling decision. The real-time replica sharding decision is used to determine the number, size, and scheme of data sharding for real-time OBD data. The real-time random encryption decision is used to determine parameters such as the random algorithm for encrypting real-time data shards and the random binary value of the random algorithm. The real-time storage resource scheduling decision is used to determine the storage resources and scheduling scheme of the storage nodes for distributed storage of data shards;

[0176] S7: The cloud data center, according to the real-time management policy, uses the blockchain network to perform distributed storage on the real-time OBD data, the corresponding target life cycle stage, and the real-time data analysis results, including the following steps:

[0177] S7-1: The cloud data center, according to the real-time replica sharding decision in the real-time management policy, performs data sharding on the real-time OBD data to obtain several real-time data shards including replica shards;

[0178] S7-2: According to the random encryption decision in the real-time management policy, use the random encryption algorithm to encrypt several real-time data shards, obtain several encrypted real-time data shards, and generate a real-time data storage request;

[0179] S7-3: Use the blockchain consensus sub-network of the blockchain network to conduct consensus on the real-time data storage request. If the consensus is successful, proceed to the next step;

[0180] S7-4: According to the storage resource scheduling decision in the real-time management policy, use the data storage sub-network of the blockchain network to perform distributed storage on several encrypted real-time data shards, and return several real-time storage addresses;

[0181] S7-5: Synchronize several real-time data shards, the corresponding real-time storage addresses, the target life cycle stage, and the real-time data analysis results to the distributed ledger.

[0182] Embodiment 2:

[0183] As Figure 2 shown, this embodiment provides a vehicle full life cycle management system based on OBD data for implementing the vehicle full life cycle management method. The system is set in the cloud data center and includes an initialization unit, a data processing unit, a model construction unit, a data classification unit, a data analysis unit, a management policy generation unit, and a distributed storage unit. The cloud data center is respectively communicatively connected to the OBD systems of several vehicles.

[0184] The OBD system is used to collect the real-time OBD data of the vehicle and upload it to the cloud data center;

[0185] The initialization unit is used to deploy the blockchain network, define several life cycle stages of the vehicle full life cycle, and collect the historical OBD data of several life cycle stages;

[0186] The data processing unit is used to perform clustering processing and key feature screening on several historical OBD data to obtain several historical OBD data clusters of life cycle stages and the corresponding key feature indicators;

[0187] The model construction unit is used to construct a data classification model, a data analysis model, and a management policy generation model according to several historical OBD data clusters and the corresponding key feature indicators by using artificial intelligence algorithms;

[0188] The data classification unit is used to receive the real-time OBD data sent by the OBD system and perform data classification on the real-time OBD data by using the data classification model to obtain the corresponding target life cycle stage;

[0189] A data analysis unit, configured to perform data analysis on real-time OBD data using a data analysis model based on target key feature indicators in a target life cycle stage, so as to obtain a real-time data analysis result;

[0190] A management strategy generation unit, configured to generate a management strategy using a management strategy generation model according to the real-time data analysis result, so as to obtain a real-time management strategy;

[0191] A distributed storage unit, configured to perform distributed storage on the real-time OBD data, the corresponding target life cycle stage, and the real-time data analysis result using a blockchain network according to the real-time management strategy.

[0192] The present invention discloses a vehicle full life cycle management method and system based on OBD data. It defines several stages of the vehicle full life cycle, and collects historical OBD data, performs data analysis, and formulates strategies for different stages, making the management strategy more targeted and effective, and improving the data management effect; it establishes a cloud data center to realize the centralized storage, management, and analysis of OBD data. Through a data sharing and exchange mechanism, it breaks data islands and realizes data integration across vehicles, brands, and regions, improving data utilization efficiency; it uses a blockchain network constructed by blockchain technology to encrypt and store OBD data to ensure the security and immutability of the data; it constructs a data classification model and a data analysis model based on artificial intelligence algorithms to achieve more accurate and comprehensive data analysis, improving the analysis ability of OBD data, ensuring the real-time nature of data management, and adjusting model parameters based on different life cycle stages to improve the accuracy of data analysis; it constructs a management strategy generation model based on artificial intelligence algorithms to dynamically generate and optimize management strategies, making the management strategy more flexible and adaptable, improving the intelligent level of data management, avoiding relying on manual experience and regular inspections, and realizing precise and efficient management.

[0193] The present invention is not limited to the above optional implementation manners. Any person can obtain other various forms of products under the inspiration of the present invention. The above specific implementation manners should not be construed as limiting the protection scope of the present invention. The protection scope of the present invention should be defined by the claims, and the specification can be used to interpret the claims.

Claims

1. A vehicle life cycle management method based on OBD data, characterized by: The steps include: Cloud data center, deploys blockchain network, defines several life cycle stages of the vehicle's entire life cycle, and collects historical OBD data of several life cycle stages; The cloud data center performs clustering processing and key feature screening on a number of historical OBD data to obtain historical OBD data clusters and corresponding key feature indicators at a number of life cycle stages; The cloud data center uses artificial intelligence algorithms to build data classification models, data analysis models, and management strategy generation models based on several historical OBD data clusters and corresponding key feature indicators; The cloud data center receives the real-time OBD data sent by the OBD system and uses the data classification model to classify the real-time OBD data to obtain the corresponding target life cycle stage; The cloud data center uses a data analysis model to analyze the real-time OBD data based on the target key characteristic indicators of the target life cycle stage to obtain real-time data analysis results; The cloud data center uses the management policy generation model to generate management policies based on the real-time data analysis results to obtain real-time management policies; The cloud data center,uses the blockchain network to perform distributed storage of real-time OBD data, the corresponding target life cycle stages, and real-time data analysis results according to the real-time management strategy.

2. A vehicle life cycle management method based on OBD data according to claim 1, characterized in that: The cloud data center deploys a blockchain network, defines several life cycle stages of the vehicle's entire life cycle, and collects historical OBD data of several life cycle stages, including the following steps: Using blockchain technology, all data servers in the cloud data center are connected as data nodes in a distributed manner to obtain a blockchain network; According to the identity allocation mechanism, several data nodes of the blockchain network are divided into several consensus nodes and several storage nodes to obtain a data storage sub-network and a blockchain consensus sub-network; Defining several life cycle stages of the vehicle's entire life cycle; the life cycle stages include the purchase stage, the running-in stage, the stabilization stage, the aging stage, and the scrapping stage; A number of historical OBD data at each life cycle stage are collected, and the number of historical OBD data are preprocessed to obtain a number of preprocessed historical OBD data belonging to different life cycle stages.

3. A vehicle life cycle management method based on OBD data according to claim 2, characterized in that: The cloud data center performs clustering processing and key feature screening on several historical OBD data to obtain historical OBD data clusters and corresponding key feature indicators at several life cycle stages, including the following steps: The cloud data center uses the AP clustering algorithm to cluster several pre-processed historical OBD data belonging to different life cycle stages and obtain several cluster centers; Set the corresponding life cycle stage for each cluster center, and obtain the Euclidean distance between each pre-processed historical OBD data and the cluster centers of all life cycle stages; All pre-processed historical OBD data are divided into cluster centers with the closest Euclidean distance to obtain historical OBD data clusters at different life cycle stages; Use the RF algorithm to obtain the importance score of each feature index of the pre-processed historical OBD data corresponding to the cluster center, and take the top several feature indexes with the highest importance scores as key feature indexes; The pre-processed historical OBD data corresponding to all cluster centers are traversed to obtain the key characteristic indicators of the historical OBD data clusters at several life cycle stages.

4. A vehicle life cycle management method based on OBD data according to claim 3, characterized in that: The cloud data center uses artificial intelligence algorithms to build data classification models, data analysis models, and management strategy generation models based on several historical OBD data clusters and corresponding key feature indicators, including the following steps: Based on several historical OBD data clusters and corresponding life cycle stages, a data classification model is constructed using a deep learning algorithm; According to the key characteristic indicators of the historical OBD data cluster, a corresponding attention weight value is set for each historical OBD data cluster; Based on several historical OBD data clusters and corresponding attention weight values, a deep learning algorithm is used to build a data analysis model and generate several historical weighted fusion features and corresponding historical data analysis results; Based on several historical weighted fusion features and corresponding historical data analysis results, a management strategy generation model is constructed using a reinforcement learning algorithm, and several historical management strategy generation experiences are generated.

5. A vehicle life cycle management method based on OBD data according to claim 4, characterized in that: The data classification model is constructed based on the LSTM-Elman algorithm, and the data classification model includes a semantic data feature extraction module constructed based on the LSTM algorithm and a data classification module constructed based on the Elman algorithm, which are connected in sequence; The data analysis model is constructed based on the DBN-Attention-MLP algorithm, and the data analysis model includes a deep feature extraction module constructed based on the DBN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the MLP algorithm, which are connected in sequence; The management strategy generation model is constructed based on the MPO-MOGRPO algorithm, and the management strategy generation model includes a meta-strategy optimization module constructed based on the MPO algorithm and a management strategy generation module constructed based on the MOGRPO algorithm. The management strategy generation module includes an objective function set, an experience replay pool, a strategy network and an intelligent agent. The intelligent agent is respectively connected to the objective function set, the experience replay pool and the strategy network, and the meta-strategy optimization module is respectively connected to the strategy network and the experience replay pool.

6. A vehicle life cycle management method based on OBD data according to claim 5, characterized in that: Based on several historical OBD data clusters and corresponding life cycle stages, a data classification model is constructed using a deep learning algorithm, including the following steps: The life cycle stage is used as the true label of several pre-processed historical OBD data in the corresponding historical OBD data cluster; According to the ratio of 7:3, the pre-processed historical OBD data with real labels of all historical OBD data clusters are divided into a training sample set and a test sample set; Use the LSTM-Elman algorithm to build an initial data classification model, input a training sample set, optimize the initial data classification model, and obtain an optimized data classification model; Input the test sample set, optimize and train the initial data classification model, and obtain several prediction labels of the optimized data classification model; Compare and count several predicted labels and the corresponding true labels to obtain the test accuracy of the optimized data classification model; If the test accuracy is greater than the accuracy threshold, the final data classification model is output, otherwise, the optimization training of the data classification model continues.

7. A vehicle life cycle management method based on OBD data according to claim 6, characterized in that: According to several historical OBD data clusters and corresponding attention weight values, a deep learning algorithm is used to build a data analysis model, and several historical weighted fusion features and corresponding historical data analysis results are generated, including the following steps: Use the DBN-Attention-MLP algorithm to build an initial data analysis model, and update the initial data analysis model according to the attention weight value of the historical OBD data cluster to obtain an updated data analysis model; Input the corresponding historical OBD data cluster, optimize and train the updated data analysis model, obtain the optimized data analysis model, and generate several historical weighted fusion features of the current historical OBD data cluster and the corresponding historical data analysis results; All historical OBD data clusters are traversed, the optimized data analysis model is optimized and trained to obtain the final data analysis model, and several historical weighted fusion features of all historical OBD data clusters and corresponding historical data analysis results are generated.

8. A vehicle life cycle management method based on OBD data according to claim 7, characterized in that: Based on several historical weighted fusion features and corresponding historical data analysis results, a management strategy generation model is constructed using a reinforcement learning algorithm, and several historical management strategy generation experiences are generated, including the following steps: Using the MPO-MOGRPO algorithm, an initial management strategy generation model is constructed; the initial management strategy generation model includes an initial meta-strategy optimization module and an initial management strategy generation module; The life cycle stage is used as the scenario for meta-strategy optimization, and the initial meta-strategy optimization module is trained based on several historical weighted fusion features under different scenarios to obtain the final meta-strategy optimization module. Using the final meta-strategy optimization module, the initial policy network of the initial management policy generation module under different scenarios is updated to obtain an updated policy network; Setting an objective function set and an experience replay pool for an initial management strategy generation module provided with an updated strategy network, and setting an action space and a state space for an agent of the initial management strategy generation module; The management strategy generation problem is used as a simulation environment, and an optimized management strategy generation module is obtained according to the updated strategy network and the intelligent agents with action space and state space. Traverse all the objective functions in the objective function set, optimize and train the optimized management strategy generation module according to the analysis results of several historical data, obtain the final management strategy generation module, and generate several historical management strategy generation experiences; The final meta-strategy optimization module and the final management strategy generation module are integrated to obtain the final management strategy generation model, and several historical management strategy generation experiences are stored in the experience replay pool.

9. A vehicle life cycle management method based on OBD data according to claim 8, characterized in that: The cloud data center uses the blockchain network to distribute the real-time OBD data, the corresponding target life cycle stages, and the real-time data analysis results according to the real-time management strategy, including the following steps: The cloud data center performs data sharding on the real-time OBD data according to the real-time management strategy to obtain a number of real-time data shards including replica shards; According to the real-time management strategy, a random encryption algorithm is used to encrypt a number of real-time data slices to obtain a number of encrypted real-time data slices, and a real-time data storage request is generated; Use the blockchain consensus sub-network of the blockchain network to reach consensus on the real-time data storage request. If the consensus is successful, proceed to the next step; According to the real-time management strategy, the data storage sub-network of the blockchain network is used to distribute and store several encrypted real-time data shards, and return several real-time storage addresses; Synchronize several real-time data shards, corresponding real-time storage addresses, target lifecycle stages, and real-time data analysis results to the distributed ledger.

10. A vehicle life cycle management system based on OBD data, used to implement the vehicle life cycle management method according to any one of claims 1 to 9, characterized in that: The system is arranged in a cloud data center, and the system includes an initialization unit, a data processing unit, a model building unit, a data classification unit, a data analysis unit, a management strategy generation unit and a distributed storage unit. The cloud data center is respectively communicated with the OBD systems of several vehicles.

Images