A motorcycle test data monitoring method and system

Abstract

The present invention relates to the technical field of test data monitoring, and specifically to a method and system for monitoring motorcycle test data. First, test data is acquired, a first influence index is obtained through a first interaction model, a second influence index is obtained through a second interaction model, and a third influence index is obtained through a third interaction model; then a comprehensive state vector is obtained based on the real-time test data and the influence indexes, and a real-time adjustment suggestion is generated; at the same time, a set of curve performance indexes and a dynamic safety threshold are calculated, the set of curve performance indexes is monitored and early warnings are given; after a test cycle, test clustering is performed according to the set of curve performance indexes and the comprehensive state vector, and vehicle parameters are adjusted according to the test clustering results. The present invention improves the accuracy of the test through effective monitoring of the data.

Description

Technical Field

[0001] The present invention relates to the technical field of test data monitoring, and particularly to a method and system for monitoring motorcycle test data. Background Art

[0002] Based on the feedback control method of multi-modal sensor fusion, the synchronous acquisition and feature extraction of vehicle dynamic parameters, environmental physical quantities, and control input signals are realized. By constructing a dynamic coupling model of engine output torque - speed - temperature and combining with the Kalman filter algorithm, the optimization of engine operating parameters is achieved; in the control of the braking system, based on the PID feedforward compensation algorithm for real-time detection of road adhesion coefficient, the prediction accuracy of braking distance is significantly improved. By using a fuzzy PID and neural network composite controller, the inertia simulation parameters of the chassis dynamometer can be dynamically adjusted, and a vehicle response prediction model under extreme conditions can be established to realize the adaptive dynamic control of the test process.

[0003] However, the existing technologies mainly focus on the data of the vehicle itself and ignore the interactive effects among the operator, vehicle state, and environmental conditions. Especially during the cornering test, the operator's behavior and posture, the parameter design of the vehicle, and environmental changes jointly affect the dynamic performance of the vehicle in the corner. The operator's control behavior directly affects the vehicle's performance in the corner, and the driving stability is also related to environmental conditions. Therefore, it is necessary to effectively monitor and accurately analyze the interactive effects of these factors in real time to safely and accurately evaluate the vehicle's performance.

[0004] Therefore, a method and system for monitoring motorcycle test data are proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for monitoring motorcycle test data. First, test data is obtained, the first influence index is obtained through the first interaction model, the second influence index is obtained through the second interaction model, and the third influence index is obtained through the third interaction model; then, based on the real-time test data and influence indexes, a comprehensive state vector is obtained, and real-time adjustment suggestions are generated; at the same time, a set of cornering performance indexes and a dynamic safety threshold are calculated, the set of cornering performance indexes is monitored and early warnings are given; after a test cycle, test clustering is performed according to the set of cornering performance indexes and the comprehensive state vector, and the vehicle parameters are adjusted according to the test clustering results. The present invention improves the accuracy of the test through effective monitoring of data.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for monitoring motorcycle test data, comprising:

[0008] Obtain real-time test data, including operator data, vehicle status data, and environmental status data; the operator data includes physical status data and operation control data;

[0009] Process the operator data and the vehicle status data through a first interaction model to obtain a first impact index, process the vehicle status data and the environmental status data through a second interaction model to obtain a second impact index, and process the operator data and the environmental status data through a third interaction model to obtain a third impact index;

[0010] Obtain a comprehensive status vector based on the real-time test data, the first impact index, the second impact index, and the third impact index, and generate real-time adjustment suggestions according to the comprehensive status vector;

[0011] Calculate a set of curve performance indicators based on the real-time test data, and calculate a dynamic safety threshold. Monitor the set of curve performance indicators according to the dynamic safety threshold and give an alarm;

[0012] After one test cycle, perform test clustering according to the set of curve performance indicators and the comprehensive status vector, and adjust the vehicle parameters according to the test clustering results.

[0013] Furthermore, the first interaction model adopts a multi-layer perceptron structure. The input layer receives the operator data and the vehicle status data. The hidden layer constructs a feature mapping through a non-linear activation function, and the output layer generates the first impact index through an activation function;

[0014] The second interaction model uses a convolutional neural network and a long short-term memory network. The input layer receives the vehicle status data and the environmental status data. The convolutional layer extracts spatial features, the LSTM layer captures time series features, the fully connected layer integrates the spatial features and the time series features, and the output layer generates the second impact index through an activation function;

[0015] The third interaction model adopts a GRU network based on an attention mechanism. The input layer receives the operator data and the environmental status data. The attention layer calculates the dynamic influence weight of environmental factors on the operator, and the GRU layer outputs the third impact index according to the dynamic influence weight.

[0016] Furthermore, define the operator, vehicle, and environment as graph nodes, use the corresponding real-time test data as the feature vectors of the graph nodes, and use the first impact index, the second impact index, and the third impact index as edge weights to construct a complete undirected graph; adopt a graph attention network, aggregate the node feature vectors through a multi-head attention mechanism to obtain an aggregated feature, and generate the comprehensive status vector through global pooling processing of the aggregated feature.

[0017] Further, an operator adjustment model is constructed based on a decision tree, and the comprehensive state vector is input into the operator adjustment model to identify the operator's adjustment requirements; the operator's adjustment requirements include steering control adjustment, speed control adjustment, attitude adjustment, and no adjustment; according to the operator's adjustment requirements, the real-time adjustment suggestions are generated through a knowledge graph; the urgency is calculated based on the change rate of the comprehensive state vector , and the reminder method for the real-time adjustment suggestions is selected according to the urgency; when , reminders are made through sound, vision, and touch; when , reminders are made through sound and vision; when , reminders are made through vision; and are the first emergency threshold and the second emergency threshold respectively.

[0018] Further, the dynamic safety threshold calculation formula for the bend performance index is:

[0019] ;

[0020] wherein, is the dynamic safety threshold, and are the basic safety threshold and the standard deviation of the bend performance index obtained from historical test data respectively, and are the current comprehensive state vector and the previous comprehensive state vector respectively, is the safety factor, is the weight coefficient of the comprehensive state vector, is the Euclidean norm.

[0021] Further, a test period is defined, and each test period includes K rounds of tests. After each round of tests, a set of bend performance indexes and a sequence of comprehensive state vectors are obtained;

[0022] According to the set of bend performance indexes and the sequence of comprehensive state vectors in a test period, a spatio-temporal tensor is constructed, and the spatio-temporal tensor is processed by a three-dimensional convolution kernel to extract correlation features;

[0023] According to the correlation features, spectral clustering is used to divide the K rounds of tests into C test abnormal patterns; the vehicle parameters corresponding to each test abnormal pattern are analyzed through a regression model respectively to guide the adjustment of vehicle parameters.

[0024] A motorcycle test data monitoring system, comprising:

[0025] A test data acquisition module, which acquires real-time test data, including operator data, vehicle state data, and environmental state data; the operator data includes physical state data and operation control data;

[0026] A data interaction analysis module processes the operator data and the vehicle status data through a first interaction model to obtain a first influence index, processes the vehicle status data and the environmental status data through a second interaction model to obtain a second influence index, and processes the operator data and the environmental status data through a third interaction model to obtain a third influence index;

[0027] A real-time adjustment suggestion module obtains a comprehensive status vector based on the real-time test data, the first influence index, the second influence index, and the third influence index, and generates real-time adjustment suggestions based on the comprehensive status vector;

[0028] A curve test analysis module calculates a set of curve performance indexes based on the real-time test data, calculates a dynamic safety threshold, monitors the set of curve performance indexes based on the dynamic safety threshold, and issues a warning;

[0029] A vehicle parameter suggestion module, after a test cycle, performs test clustering based on the set of curve performance indexes and the comprehensive status vector, and adjusts vehicle parameters according to the test clustering results.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] 1. By using three interaction models to process the dynamic interaction relationships between the operator-vehicle, vehicle-environment, and operator-environment respectively, and then conducting a joint analysis of the three as a whole. The first interaction model captures the non-linear relationship between the operator and the vehicle; the second interaction model extracts the spatio-temporal correlation between the vehicle and the environment; the third interaction model dynamically quantifies the influence weight of the environment on the operator, avoiding the limitations of a single model. Further obtaining a comprehensive status vector based on the real-time test data and influence indexes provides a multi-dimensional basis for subsequent decision-making. Through the interaction analysis of the three-source data, the characterization ability of the test data is comprehensively improved, realizing the comprehensive monitoring of the human-vehicle-environment interaction effect during the test process, and providing more accurate inputs for real-time adjustment suggestions and safety monitoring.

[0032] 2. At the real-time level, based on a decision tree and knowledge graph-based operator adjustment model, combined with the short-term change rate of the comprehensive status vector to calculate the urgency, dynamically generate multi-modal reminders to ensure that the operator responds to risks in a timely manner. At the periodic level, spatio-temporal correlation features of multiple rounds of tests are extracted through spatio-temporal tensors, and three-dimensional convolution and spectral clustering are used to identify test abnormal patterns, associate vehicle parameter defects, and generate adjustment suggestions. Real-time intervention solves sudden risks, while periodic optimization helps to correct the shortcomings of vehicle design. The cooperation of the two not only ensures test safety but also improves test effectiveness, achieving the dual goals of dynamic risk response and continuous vehicle optimization.

[0033] 3. Define the operator, vehicle, and environment as graph nodes, use real-time data as node features, and use three impact indicators as edge weights to construct a complete undirected graph. Aggregate node features through the multi-head attention mechanism of the graph attention network, and then generate a comprehensive state vector through global pooling. This can capture implicit associations while retaining local interaction details and global state features, enabling the comprehensive state vector to have both high-dimensional representation capabilities and robustness, providing a unified high-quality input for real-time decision-making and periodic optimization. The application of the graph attention network realizes the efficient fusion of multi-source data, providing a more comprehensive state representation for real-time adjustment suggestions and long-term parameter optimization during the testing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Schematic flowchart of a method for monitoring motorcycle test data according to the present invention;

[0035] Figure 2 Schematic structural diagram of a motorcycle test data monitoring system according to the present invention;

[0036] Figure 3 Schematic flowchart of obtaining data for the comprehensive state vector according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] Please refer to Figures 1 to 3 , the present invention provides a method and system for monitoring motorcycle test data, and the technical solutions are as follows: Embodiment

[0039] To evaluate and optimize the performance of a motorcycle during cornering, as Figure 1 shown, a method for monitoring motorcycle test data is applied, including:

[0040] Obtain real-time test data, including operator data, vehicle state data, and environmental state data; the operator data includes physical state data and operation control data;

[0041] Process the operator data and the vehicle state data through a first interaction model to obtain a first impact indicator, process the vehicle state data and the environmental state data through a second interaction model to obtain a second impact indicator, and process the operator data and the environmental state data through a third interaction model to obtain a third impact indicator;

[0042] Obtain a comprehensive state vector based on the real-time test data, the first impact index, the second impact index, and the third impact index, and generate real-time adjustment suggestions according to the comprehensive state vector;

[0043] Calculate a set of curve performance indicators based on the real-time test data, and calculate a dynamic safety threshold. Monitor the set of curve performance indicators according to the dynamic safety threshold and give early warnings;

[0044] After one test cycle, perform test clustering according to the set of curve performance indicators and the comprehensive state vector, and adjust the vehicle parameters according to the test clustering results.

[0045] Table 1 shows the acquisition method of some original test data. The original test data realizes multi-frequency data synchronization through timestamp alignment and interpolation algorithms. For high-frequency data, a sliding window is used for downsampling to the lowest common frequency to ensure temporal consistency; for low-frequency data, the intermediate time points are filled by linear interpolation. After the data flow is denoised and normalized, a real-time test data sequence is generated.

[0046] Table 1 Original Test Data Acquisition

[0047]

[0048] The interactive effects among the operator, vehicle, and environmental data are accurately quantified through three interactive models, and a comprehensive state vector is generated by combining a graph attention network, realizing comprehensive monitoring of the test process. Real-time adjustment suggestions are generated based on the comprehensive state vector, and different reminder methods are selected according to the urgency to ensure test safety. At the same time, a dynamic safety threshold mechanism is introduced to enable safety monitoring to adapt to complex and changing test environments and identify potential risks in advance. In addition, through the clustering analysis of the test cycle, the system can discover regular problems in the test and provide a scientific basis for vehicle parameter optimization.

[0049] Furthermore, the first interactive model adopts a multi-layer perceptron structure. The input layer receives the operator data and the vehicle state data, the hidden layer constructs a feature mapping through a non-linear activation function, and the output layer generates the first impact index through an activation function;

[0050] The second interactive model uses a convolutional neural network and a long short-term memory network. The input layer receives the vehicle state data and the environmental state data, the convolutional layer extracts spatial features, the LSTM layer captures time series features, the fully connected layer integrates the spatial features and the time series features, and the output layer generates the second impact index through an activation function;

[0051] The third interaction model uses a GRU network based on the attention mechanism. The input layer receives the operator data and the environmental state data. The attention layer calculates the dynamic influence weight of environmental factors on the operator. The GRU layer outputs the third influence index according to the dynamic influence weight.

[0052] By using a multi-layer perceptron, a convolutional neural network combined with a long short-term memory network, and a GRU network based on the attention mechanism to process different-dimensional interaction relationships respectively, it is possible to accurately capture the complex non-linear relationships between different types of data, ensure targeted capture of key influencing factors, comprehensively understand the interaction effects among the operator-vehicle-environment, and improve the accuracy and reliability of the interaction influence index.

[0053] Furthermore, the operator, the vehicle, and the environment are defined as graph nodes, and the corresponding real-time test data is used as the feature vector of the graph node. The first influence index, the second influence index, and the third influence index are used as edge weights to construct a complete undirected graph. A graph attention network is adopted, and the node feature vectors are aggregated through a multi-head attention mechanism to obtain an aggregated feature, and the aggregated feature is processed by global pooling to generate the comprehensive state vector.

[0054] The graph structure and the graph attention network are introduced to process the comprehensive state. The operator, the vehicle, and the environment are defined as graph nodes, and a complete undirected graph is constructed with the three influence indexes as edge weights. The node feature vectors are aggregated through a multi-head attention mechanism and global pooling is performed to generate the comprehensive state vector. The graph-based representation method can capture the complex associations among various elements during the motorcycle test, provide a more comprehensive and accurate state representation, and lay a solid foundation for the generation of subsequent adjustment suggestions.

[0055] Furthermore, an operator adjustment model is constructed based on a decision tree. The comprehensive state vector is input into the operator adjustment model to identify the operator's adjustment needs. The operator's adjustment needs include steering control adjustment, speed control adjustment, attitude adjustment, and no adjustment. According to the operator's adjustment needs, the real-time adjustment suggestions are generated through a knowledge graph. The urgency is calculated based on the change rate of the comprehensive state vector , and the reminder method for the real-time adjustment suggestions is selected according to the urgency; when , reminders are made through sound, vision, and touch; when , reminders are made through sound and vision; when , reminders are made through vision; and are the first emergency threshold and the second emergency threshold respectively.

[0056] Table 2 shows the example data of real-time adjustment suggestions. An operator adjustment model is constructed through the random forest algorithm, and the training data is sourced from historical test data and the best practice experience annotated by professional drivers.

[0057] Table 2 Example Data of Real-Time Adjustment Suggestions

[0058]

[0059] The generation of adjustment suggestions based on decision trees and knowledge graphs improves the safety during the testing process. Combined with the urgency grading reminder, it can provide feedback of appropriate intensity, avoiding over-intervention in low-risk situations and timely and effectively reminding the operator in high-risk situations, ensuring that the operator receives the most suitable feedback at different risk levels and reducing misjudgment and response delay.

[0060] Furthermore, the dynamic safety threshold calculation formula for the corner performance index is:

[0061] ;

[0062] where is the dynamic safety threshold, and are respectively the basic safety threshold and the standard deviation of the corner performance index obtained from historical test data, and are respectively the current comprehensive state vector and the previous comprehensive state vector, is the safety factor, is the weight coefficient of the comprehensive state vector, is the Euclidean norm.

[0063] Furthermore, the corner performance index includes the corner deceleration control index , the corner stability index , the corner acceleration smoothness index and the grip utilization index ; the calculation formulas are:

[0064] ; ; ; ;

[0065] where , and are respectively the actual deceleration, the ideal deceleration and the maximum safe deceleration, is the minimum value function; and are respectively the standard deviation of the inclination angle in the corner and the maximum safe inclination angle; and are the standard deviation of acceleration and the maximum safe acceleration respectively; , , and are the longitudinal acceleration, lateral acceleration, road surface friction coefficient and gravitational acceleration respectively.

[0066] In this embodiment, a comprehensive state vector is obtained by setting a time window. Every other time window, the first influence index, the second influence index and the third influence index are obtained according to the real-time test data within the cumulative time window, and the comprehensive state vector is obtained through a graph attention network in combination with the current real-time test data; the cornering deceleration control index, the cornering stability index and the cornering acceleration smoothness index in the cornering performance index are the performance indexes corresponding to the cornering entrance stage, the cornering stage and the cornering exit stage respectively. By dividing the cornering test site into a cornering entrance area, a cornering area and a cornering exit area, and combining the sensor to sense the area where the motorcycle is located, the corresponding performance indexes are calculated, and the grip utilization rate index needs to be calculated in all areas; therefore, when monitoring the cornering performance index, the performance indexes and the grip utilization rate index of the corresponding area are monitored through the dynamic safety threshold corresponding to each performance index.

[0067] Combining the basic safety threshold with the change of the comprehensive state vector realizes the adaptive adjustment of the threshold. The dynamic threshold mechanism takes into account the statistical characteristics of historical data and the real-time changes of the current test environment, making the safety monitoring more flexible and accurate, solving the lag problem of fixed thresholds in complex scenarios, and improving the effectiveness of safety monitoring.

[0068] Furthermore, a test period is defined. Each test period includes K rounds of tests. After each round of tests, a set of cornering performance indexes and a sequence of comprehensive state vectors are obtained;

[0069] A spatio-temporal tensor is constructed according to the set of cornering performance indexes and the sequence of comprehensive state vectors in a test period, and the associated features are extracted by processing the spatio-temporal tensor with a three-dimensional convolution kernel;

[0070] According to the associated features, spectral clustering is used to divide the K rounds of tests into C types of test anomaly patterns; the vehicle parameters corresponding to each test anomaly pattern are analyzed through a regression model respectively to guide the adjustment of vehicle parameters.

[0071] Specifically, each test period includes 10 rounds of tests. After each round of tests, a set of cornering performance indexes and a sequence of comprehensive state vectors are collected, and a spatio-temporal tensor is constructed through outer product operation. Table 3 gives the test clustering results obtained in multiple test periods, which are used to guide the adjustment of vehicle parameters.

[0072] Table 3 Test Clustering Results

[0073]

[0074] The clustering analysis method based on the test cycle constructs a spatio-temporal tensor, extracts correlation features, and uses spectral clustering to identify test anomaly patterns, providing a basis for vehicle parameter adjustment. The periodic analysis method can discover potential laws and systematic problems in the test process. Through the comprehensive analysis of multi-round test data, the system can identify the key parameters affecting vehicle performance under specific conditions and provide a more targeted basis for vehicle parameter adjustment. Embodiment

[0075] This embodiment provides a motorcycle test data monitoring system. The system structure is shown in Figure 2 , including a test data acquisition module, a data interaction and analysis module, a real-time adjustment suggestion module, and a vehicle parameter suggestion module, and is used to implement the following steps:

[0076] Obtain real-time test data, including operator data, vehicle status data, and environmental status data; the operator data includes physical status data and operation control data;

[0077] As Figure 3 shown, the first interaction model processes the operator data and the vehicle status data to obtain a first influence index, the second interaction model processes the vehicle status data and the environmental status data to obtain a second influence index, and the third interaction model processes the operator data and the environmental status data to obtain a third influence index;

[0078] As Figure 3 shown, a comprehensive state vector is obtained according to the real-time test data, the first influence index, the second influence index, and the third influence index, and a real-time adjustment suggestion is generated according to the comprehensive state vector;

[0079] The corner test analysis module calculates a set of corner performance indicators according to the real-time test data, calculates a dynamic safety threshold, monitors the set of corner performance indicators according to the dynamic safety threshold, and issues a warning;

[0080] After a test cycle, test clustering is performed according to the set of corner performance indicators and the comprehensive state vector, and vehicle parameters are adjusted according to the test clustering results.

[0081] Furthermore, the first interaction model adopts a multi-layer perceptron structure. The input layer receives the operator data and the vehicle status data, the hidden layer constructs a feature map through a non-linear activation function, and the output layer generates the first influence index through an activation function;

[0082] The second interaction model uses a convolutional neural network and a long short-term memory network. The input layer receives the vehicle state data and the environmental state data. The convolutional layer extracts spatial features, the LSTM layer captures time series features, the fully connected layer integrates the spatial features and the time series features, and the output layer generates the second influence index through an activation function;

[0083] The third interaction model adopts a GRU network based on an attention mechanism. The input layer receives the operator data and the environmental state data. The attention layer calculates the dynamic influence weight of environmental factors on the operator, and the GRU layer outputs the third influence index according to the dynamic influence weight.

[0084] Further, the operator, the vehicle, and the environment are defined as graph nodes, the corresponding real-time test data is used as the feature vector of the graph node, and the first influence index, the second influence index, and the third influence index are used as edge weights to construct a complete undirected graph; a graph attention network is adopted, and the node feature vectors are aggregated through a multi-head attention mechanism to obtain an aggregated feature, and the aggregated feature is processed through global pooling to generate the comprehensive state vector.

[0085] Further, an operator adjustment model is constructed based on a decision tree, and the comprehensive state vector is input into the operator adjustment model to identify the operator adjustment requirements; the operator adjustment requirements include steering control adjustment, speed control adjustment, attitude adjustment, and no adjustment; according to the operator adjustment requirements, the real-time adjustment suggestions are generated through a knowledge graph; the urgency is calculated based on the change rate of the comprehensive state vector , and the reminder method of the real-time adjustment suggestion is selected according to the urgency; when , reminders are made through sound, vision, and touch; when , reminders are made through sound and vision; when , reminders are made through vision; and are the first emergency threshold and the second emergency threshold respectively.

[0086] Further, the dynamic safety threshold calculation formula for the curve performance index is:

[0087] ;

[0088] Among them, is the dynamic safety threshold, and are the basic safety threshold and the standard deviation of the curve performance index obtained from historical test data respectively, and are the current comprehensive state vector and the previous comprehensive state vector respectively, is the safety factor, is the weight coefficient of the comprehensive state vector, is the Euclidean norm.

[0089] Furthermore, a test period is defined. Each test period includes K rounds of tests. After each round of tests, a set of cornering performance indicators and a sequence of comprehensive state vectors are obtained;

[0090] A spatio-temporal tensor is constructed based on the set of cornering performance indicators and the sequence of comprehensive state vectors in a test period, and the associated features are extracted by processing the spatio-temporal tensor with a three-dimensional convolution kernel;

[0091] Based on the associated features, spectral clustering is used to divide the K rounds of tests into C types of test anomaly patterns; the vehicle parameters corresponding to each test anomaly pattern are analyzed through a regression model respectively to guide the adjustment of vehicle parameters.

[0092] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A motorcycle test data monitoring method, characterized in that: include: Acquire real-time test data, including operator data, vehicle status data and environmental status data; operator data includes physical status data and operation control data; Processing the operator data and the vehicle status data through the first interaction model to obtain a first impact index, processing the vehicle status data and the environment status data through the second interaction model to obtain a second impact index, and processing the operator data and the environment status data through the third interaction model to obtain a third impact index; The first interactive model adopts a multi-layer perceptron structure. The input layer receives operator data and vehicle status data, the hidden layer constructs a feature map through a nonlinear activation function, and the output layer generates a first impact indicator through an activation function. The second interactive model uses a convolutional neural network and a long short-term memory network. The input layer receives vehicle status data and environmental status data, the convolution layer extracts spatial features, the LSTM layer captures time series features, the fully connected layer integrates spatial features and time series features, and the output layer generates a second impact indicator through an activation function. The third interactive model adopts a GRU network based on an attention mechanism. The input layer receives operator data and environmental status data, the attention layer calculates the dynamic impact weight of environmental factors on the operator, and the GRU layer outputs the third impact indicator based on the dynamic impact weight. A comprehensive state vector is obtained according to the real-time test data, the first influencing index, the second influencing index and the third influencing index, and a real-time adjustment suggestion is generated according to the comprehensive state vector, including: building an operator adjustment model based on a decision tree, inputting the comprehensive state vector into the operator adjustment model, and identifying the operator adjustment requirements; the operator adjustment requirements include steering control adjustment, speed control adjustment, posture adjustment and no adjustment required; and generating real-time adjustment suggestions according to the operator adjustment requirements through a knowledge graph; Calculate the set of cornering performance indicators based on real-time test data, calculate the dynamic safety threshold, monitor the set of cornering performance indicators based on the dynamic safety threshold and issue an early warning; After a test cycle, test clustering is performed based on the set of cornering performance indicators and the comprehensive state vector, and vehicle parameters are adjusted based on the test clustering results.

2. A motorcycle test data monitoring method according to claim 1, characterized in that: The operator, vehicle and environment are defined as graph nodes, the corresponding real-time test data are used as the feature vectors of the graph nodes, the first influencing indicator, the second influencing indicator and the third influencing indicator are used as edge weights, and a completely undirected graph is constructed; a graph attention network is adopted, and the node feature vectors are aggregated through a multi-head attention mechanism to obtain aggregated features, and the aggregated features are processed by global pooling to generate the comprehensive state vector.

3. A motorcycle test data monitoring method according to claim 1, characterized in that: Calculate the urgency based on the comprehensive state vector change rate , select the reminder method of the real-time adjustment suggestion according to the urgency; when When When When the time comes, use visual reminders; and They are the first emergency threshold and the second emergency threshold respectively.

4. A motorcycle test data monitoring method according to claim 1, characterized in that: The calculation formula of the dynamic safety threshold of the cornering performance index is: ； in, is the dynamic safety threshold, and are the basic safety threshold and standard deviation of the cornering performance index obtained based on historical test data, and are the current integrated state vector and the previous integrated state vector respectively, is the safety factor, is the weight coefficient of the comprehensive state vector, is the Euclidean norm.

5. A motorcycle test data monitoring method according to claim 1, characterized in that: Define a test cycle. Each test cycle includes K rounds of testing. After each round of testing, a set of cornering performance indicators and a comprehensive state vector sequence are obtained. Constructing a space-time tensor according to a set of curve performance indicators and a comprehensive state vector sequence of a test cycle, and extracting correlation features by processing the space-time tensor through a three-dimensional convolution kernel; According to the correlation features, spectral clustering is used to divide the K rounds of tests into C test abnormality modes; the corresponding vehicle parameters under each test abnormality mode are analyzed by regression models to guide vehicle parameter adjustment.

6. A motorcycle test data monitoring system, characterized in that: include: A test data acquisition module acquires real-time test data, including operator data, vehicle status data, and environmental status data; operator data includes physical status data and operation control data; The data interaction analysis module processes the operator data and vehicle status data through the first interaction model to obtain the first impact index, processes the vehicle status data and environment status data through the second interaction model to obtain the second impact index, and processes the operator data and environment status data through the third interaction model to obtain the third impact index; the first interaction model adopts a multi-layer perceptron structure, the input layer receives the operator data and vehicle status data, the hidden layer constructs a feature map through a nonlinear activation function, and the output layer generates the first impact index through an activation function; the second interaction model uses a convolutional neural network and a long short-term memory network, the input layer receives the vehicle status data and environment status data, the convolution layer extracts spatial features, the LSTM layer captures time series features, the fully connected layer integrates spatial features and time series features, and the output layer generates the second impact index through an activation function; the third interaction model adopts a GRU network based on an attention mechanism, the input layer receives the operator data and environment status data, the attention layer calculates the dynamic impact weight of environmental factors on the operator, and the GRU layer outputs the third impact index according to the dynamic impact weight; The real-time adjustment suggestion module obtains a comprehensive state vector according to the real-time test data, the first influencing index, the second influencing index and the third influencing index, and generates real-time adjustment suggestions according to the comprehensive state vector, including: building an operator adjustment model based on a decision tree, inputting the comprehensive state vector into the operator adjustment model, and identifying the operator adjustment requirements; the operator adjustment requirements include steering control adjustment, speed control adjustment, posture adjustment and no adjustment; and generating real-time adjustment suggestions through a knowledge graph according to the operator adjustment requirements; The curve test analysis module calculates the curve performance index set based on the real-time test data, calculates the dynamic safety threshold, monitors the curve performance index set based on the dynamic safety threshold, and issues warnings; The vehicle parameter recommendation module, after a test cycle, performs test clustering based on the set of cornering performance indicators and the comprehensive state vector, and adjusts the vehicle parameters based on the test clustering results.

Images