A vehicle testing method for loose chassis fasteners of suspension and brake systems
The data of the chassis fastener is obtained through static and dynamic testing, combined with multi-layer perceptron and tree structure, the problem of loose chassis fasteners is identified, and the problem of insufficient detection complexity and data security in the existing technology is solved, and the problem of insufficient detection is achieved. Assessment of vehicle stability and personalized services.
Patent Information
- Application Number
- CN202411184027.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-27
AI Technical Summary
The prior art is difficult to accurately detect whether the chassis fastener is loose under dynamic or changing loading conditions, and the user data access permissions are insufficient, which affects the security of the test data and personalized services.
Static and dynamic tests are used to obtain the torque value and vibration data of the fastener, calculate the daily vibration consistency index, time period vibration consistency index and vibration stability change index, combine the vehicle motion characteristics to generate a comprehensive activity characterization vector, and use a multi-layer perception machine Identify loose problems and manage data access through tree structures and deep learning algorithms.
It provides a quantitative assessment of vehicle operation stability, identify potential fastener loosening problems, improves data security and personalized experience, realizes intelligent permission adjustment, and supports scientific maintenance decisions.
Smart Images

Figure CN118687873B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of fastener detection, and in particular to a whole vehicle testing method for loose fasteners of a suspension and brake system chassis. Background Art
[0002] The suspension system is a key part of the vehicle chassis. It connects the wheels to the vehicle body and is responsible for absorbing the impact and vibration caused by uneven roads, thereby ensuring vehicle stability and ride comfort. The main components of the suspension system include springs, shock absorbers, control arms, stabilizer bars, etc. The braking system is one of the key systems to ensure vehicle safety. It allows the driver to control the speed of the vehicle and stop when necessary. The core components of the braking system include brake shoes, brake discs, brake calipers, brake lines, and brake fluid.
[0003] Chassis fasteners include bolts, nuts and other fixings that connect suspension, brake systems and other chassis components. These fasteners are subjected to gravity, vibration and thermal stress during vehicle operation and may gradually loosen. Loosening of fasteners may increase the relative movement between components, thereby accelerating wear, generating abnormal noise, reducing vehicle handling performance, and even causing components to fall off, seriously threatening driving safety.
[0004] In order to solve the problem of loose chassis fasteners, it is necessary to include fastener torque inspection, vibration analysis and dynamic driving testing to ensure the safety and performance of the vehicle. In the prior art, for example, Chinese Patent 202110454179.8 discloses a fastener loosening detection method, device, electronic device and storage medium, which determines the loose state of the fastener based on the acceleration curve, upper limit acceleration curve and lower limit acceleration curve of the second component. However, the above method still has the following shortcomings: this method is suitable for certain specific types of fasteners and application scenarios. Under dynamic or changing loading conditions, it is more complicated to determine whether the fastener is loose, and more data and analysis may be required to ensure accuracy. At the same time, there is no setting for the user's data access rights, which cannot improve the security of fastener test data and the quality of personalized services.
[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention
[0006] In view of the problems in the related art, the present invention proposes a whole vehicle testing method for loose chassis fasteners of a suspension and braking system to overcome the above-mentioned technical problems existing in the existing related art.
[0007] To this end, the specific technical solution adopted by the present invention is as follows:
[0008] A whole vehicle testing method for loose fasteners of a suspension and brake system chassis, the whole vehicle testing method for loose fasteners of a suspension and brake system chassis comprising the following steps:
[0009] S1. Obtain the initial status of all fasteners.
[0010] S2. Perform static testing on the vehicle and obtain torque values of fasteners of the suspension system and the brake system, and perform dynamic testing on the vehicle to obtain vibration data of the vehicle, and construct test data based on the torque values and vibration data of the fasteners.
[0011] S3. Calculate the daily vibration consistency index, time period vibration consistency index and vibration stability change index of the vibration data.
[0012] S4. Based on the daily vibration consistency index, the period vibration consistency index and the vibration stability change index and integrating the vehicle's motion characteristics, a comprehensive activity representation vector and an implicit activity representation vector are calculated.
[0013] S5. Use the tree structure to analyze the test data and automatically generate a test report. At the same time, input the implicit activity representation vector into the trained multi-layer perceptron, and use the trained multi-layer perceptron to make predictions to identify whether there is a problem of fastener looseness.
[0014] S6. Based on the test report and prediction results, determine whether the fasteners are loose. If loose, record the location and degree of the loose fasteners.
[0015] Further, calculating the daily vibration consistency index, the period vibration consistency index and the vibration stability change index of the vibration data includes the following steps:
[0016] S31, calculating the entropy value of the vibration data within one day to obtain a daily vibration consistency index, and using the daily vibration consistency index to evaluate the stability of the vehicle in daily operation;
[0017] S32, by calculating the entropy value of the vibration data in different time periods, obtaining the time period vibration consistency index, and evaluating the running stability of the vehicle in each time period by the time period vibration consistency index;
[0018] S33. Obtain a vibration stability change index by comparing the entropy value changes of the vibration data in two time periods, and use the vibration stability change index to represent the change trend of the vehicle operation stability.
[0019] Furthermore, when calculating the entropy value of vibration data within a day, the entropy value calculation formula is:
[0020] ;
[0021] In the formula, HRepresents the entropy value of vibration data;
[0022] p ( x i ) indicates the i The probability of a data point;
[0023] x i Indicates i The value of the data point, n Represents the total number of data points.
[0024] Furthermore, based on the daily vibration consistency index, the period vibration consistency index and the vibration stability change index and integrating the vehicle's motion characteristics, calculating the comprehensive activity representation vector and the implicit activity representation vector includes the following steps:
[0025] S41, obtaining motion features of the vehicle, and converting the motion features into auxiliary representation vectors;
[0026] S42, fusing the vibration entropy vector including the daily vibration consistency index, the period vibration consistency index and the vibration stability change index of the vibration data with the auxiliary representation vector to obtain a fused activity representation vector;
[0027] S43. After disrupting the data sequence for multiple times, recalculate the auxiliary representation vector and the vibration entropy vector, and fuse them again to generate a new fused activity representation vector. At the same time, input the new fused activity representation vector into the encoder-decoder model, and output the implicit activity representation vector.
[0028] Furthermore, using the tree structure to analyze the test data and automatically generate a test report includes the following steps:
[0029] S51, obtaining user behavior log data, preprocessing the user behavior log data and test data, and grouping them according to vehicle type, test type and test date;
[0030] S52, constructing a tree structure, and using the grouping information as different nodes;
[0031] S53, setting data access permissions according to the user's role and identity information, and associating the user's access permissions with corresponding nodes in the tree structure, so that the user can only access data that he or she has permission to view;
[0032] S54. Analyze user behavior log data using deep learning algorithms, predict user data access needs, and dynamically adjust access permissions;
[0033] S55. Integrate the test data of each node, analyze the relationship between nodes, and identify trends and patterns;
[0034] S56. Automatically obtain test reports based on tree structure and analysis results;
[0035] S57. Customize the test report content according to the user's access rights to ensure that the user obtains the information he or she is entitled to access.
[0036] Furthermore, using deep learning algorithms to analyze user behavior log data, predicting user data access needs, and dynamically adjusting access permissions includes the following steps:
[0037] S541, obtaining co-occurrence frequency data of user behaviors from user behavior log data, and calculating the behavior similarity between users based on the co-occurrence frequency data of user behaviors to obtain a user behavior similarity matrix;
[0038] S542, inputting the user behavior similarity matrix into the sparse autoencoder, extracting the key feature representation of the user behavior, and reconstructing the user behavior similarity matrix from the hidden layer features, and adding a sparsity penalty to obtain the user behavior feature representation;
[0039] S543, determining the number of clusters, and determining the initial cluster center based on the principle of maximizing the modularity of user behavior and combining the density and similarity of user behavior features;
[0040] S544, iteratively optimizing the clustering results until a termination condition is met or a maximum number of iterations is reached, thereby obtaining a user group division result;
[0041] S545, calculating the user behavior modularity, evaluating the consistency within the group and the distinction between the groups, and adjusting the user access rights according to the user behavior modularity value;
[0042] S546. Implement the user's access rights, monitor the user's behavior changes after the access rights are adjusted, and dynamically adjust the user's access rights according to the changes in the user's behavior.
[0043] Furthermore, the number of clusters is determined, and based on the principle of maximizing the modularity of user behavior, the initial cluster center is determined in combination with the density and similarity of user behavior features, including the following steps:
[0044] S5431, obtaining a user behavior modularity function, and calculating the user behavior modularity values under different numbers of clusters through an iterative process. If the user behavior modularity value reaches the maximum, the corresponding number of clusters is the optimal number of clusters;
[0045] S5432, calculating the density of each user behavior data point within a specific radius, and sorting them from high to low according to density, and selecting the data point with the largest density as the first cluster center;
[0046] S5433, calculating the Euclidean distance and density between the remaining data points and the determined cluster center, and presetting the density threshold and distance threshold, and selecting the data points that are both greater than or equal to the density threshold and whose distance from the existing cluster center is greater than or equal to the distance threshold as the new cluster center;
[0047] S5434. Repeat S5432-S5433 until all initial cluster centers are determined.
[0048] Furthermore, the user behavior modularity function is obtained, and through an iterative process, the user behavior modularity values under different numbers of clusters are calculated. The optimization algorithm is used to maximize the user behavior modularity as the objective function, and the cluster center is adjusted in the iterative process until the cluster center configuration that can maximize the user behavior modularity is found.
[0049] Furthermore, the implicit activity representation vector is input into the trained multi-layer perceptron, and prediction is performed by the trained multi-layer perceptron to identify whether there is a fastener loosening problem, including the following steps:
[0050] Collect implicit activity representation vectors generated by previous vibration data in advance and convert them into standardized feature vectors;
[0051] Use the standardized feature vector to train and evaluate the multilayer perceptron to obtain a trained multilayer perceptron;
[0052] The newly acquired implicit activity representation vector is converted into a new format and then fed into the trained multi-layer perceptron, which then outputs a prediction result indicating whether the fastener is loose.
[0053] Furthermore, the training and evaluation of the multilayer perceptron using the standardized feature vector includes the following steps:
[0054] Assume the structure of a multilayer perceptron, including input layer, hidden layer and output layer, and select activation function and initialization weights;
[0055] Divide the standardized feature vector into training data set and test data set;
[0056] Using the gradient descent method, the multilayer perceptron is trained using the training data set, and the neuron weights are adjusted to minimize the squared error;
[0057] Evaluate the performance of the Multilayer Perceptron using the test dataset.
[0058] The beneficial effects of the present invention are:
[0059] (1) The present invention provides a vehicle test method for loose chassis fasteners of suspension and brake systems. Through static and dynamic tests, torque values and vibration data are obtained, and a comprehensive test data set is jointly constructed for subsequent analysis. By calculating the daily vibration consistency index, period vibration consistency index and vibration stability change index of vibration data, a quantitative evaluation of the daily operation stability of the vehicle is provided, and the necessary input data is provided for the calculation of the comprehensive activity representation vector. The vibration data and vehicle motion characteristics are integrated to provide a comprehensive overview of vehicle activities and generate a comprehensive data representation for a multi-layer perceptron.
[0060] (2) The tree structure is used to analyze the test data, provide a structured test report, and realize the dynamic management of data access rights. It also provides a personalized test report. The combination of rights management and user behavior analysis improves data security and personalized experience. The deep learning algorithm is used to analyze user behavior, realize the intelligent adjustment of access rights, and the calculation of user behavior modularity provides a basis for clustering and rights management.
[0061] (3) Multilayer perceptron learning is used to identify patterns of fastener loosening from historical data, and this knowledge is applied to new data to predict and diagnose potential problems. Potential fastener loosening problems are identified through multilayer perceptron learning, and the prediction results provide a scientific basis for maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0063] Figure 1 The present invention is a flow chart of a method for testing the looseness of chassis fasteners of a suspension and brake system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0064] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.
[0065] According to an embodiment of the present invention, a whole vehicle testing method for loosening chassis fasteners of a suspension and brake system is provided.
[0066] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the whole vehicle testing method for loose chassis fasteners of the suspension and brake system according to the embodiment of the present invention, the whole vehicle testing method for loose chassis fasteners of the suspension and brake system comprises the following steps:
[0067] S1. Perform a preliminary inspection of the vehicle and ensure that the vehicle is in normal condition, including normal tire pressure and no obvious damage, including checking the exterior and interior of the vehicle to ensure that there is no damage or wear that may affect the accuracy of the test results.
[0068] Check the appearance of the vehicle suspension and brake system fasteners for signs of looseness or wear, including checking the bolts for any signs of looseness, corrosion or damage. Obtain the initial state of all fasteners and measure their torque values to assess their tightness. All subsequent tests are based on a known reference point, that is, all subsequent tests are performed under the same conditions to ensure consistency and comparability of test results.
[0069] S2. Perform static testing on the vehicle and obtain torque values of fasteners of the suspension system and the brake system, and perform dynamic testing on the vehicle to obtain vibration data of the vehicle, and construct test data based on the torque values and vibration data of the fasteners.
[0070] Specifically, static testing:
[0071] Use a torque wrench to measure the torque values of the suspension and brake system fasteners. Record the torque value of each fastener.
[0072] Dynamic testing:
[0073] Use vibration sensors to collect vibration data of the vehicle during operation. Ensure that data is collected under different operating conditions (e.g., different speeds, different road surfaces).
[0074] Data Records:
[0075] The collected torque values and vibration data are recorded and annotated to ensure the accuracy and completeness of the data.
[0076] S3. Calculate the daily vibration consistency index, time period vibration consistency index and vibration stability change index of the vibration data.
[0077] In a further embodiment, calculating the daily vibration consistency index, the period vibration consistency index and the vibration stability change index of the vibration data comprises the following steps:
[0078] S31, calculating the entropy value of the vibration data within one day to obtain the daily vibration consistency index, and using the daily vibration consistency index to evaluate the stability of the vehicle in daily operation, wherein a higher entropy value indicates a more chaotic distribution of the vibration data, reflecting potential instability;
[0079] The specific calculation steps of the entropy value of vibration data are as follows:
[0080] Collect vibration data from the vehicle over a day.
[0081] The vibration data distribution of the vehicle within a day is normalized so that the sum of all values is 1.
[0082] Use the entropy formula to calculate the entropy value of the data.
[0083] S32. By calculating the entropy value of vibration data in different time periods, such as morning and afternoon, the calculation of the daily vibration consistency index is repeated for the vibration data in each time period to obtain the time period vibration consistency index, and the operation stability of the vehicle in each time period is evaluated by the time period vibration consistency index; it is helpful to identify whether there is abnormal vibration in a specific time period, so as to carry out targeted maintenance.
[0084] S33. By comparing the entropy changes of vibration data for two consecutive days or two periods, that is, calculating the difference or ratio between the daily vibration consistency indexes of the two periods, the vibration stability change index is obtained, and the vibration stability change index is used to represent the change trend of vehicle operation stability; it is helpful to predict possible problems in the future so that maintenance can be carried out in time.
[0085] In a further embodiment, when calculating the entropy value of vibration data within a day, the probability of each data point is first determined, and then the logarithm of each probability is calculated, and then the products are added, and finally the inverse of the sum is taken to obtain the entropy value. The higher the entropy value, the greater the uncertainty of the data; the lower the entropy value, the higher the consistency of the data, which helps to evaluate the stability of the vehicle's vibration data within a certain period of time, so as to determine whether there is a loose problem with the fasteners of the suspension and braking system. The entropy value calculation formula is:
[0086] ;
[0087] In the formula, H The entropy value representing the vibration data is a measure of the randomness or uncertainty of the data;
[0088] p ( x i ) indicates the i The probability of a data point, that is, the relative frequency of the data point in all the data;
[0089] x i Indicates i The value of the data point, n Represents the total number of data points.
[0090] S4. Based on the daily vibration consistency index, the period vibration consistency index and the vibration stability change index and integrating the vehicle's motion characteristics, a comprehensive activity representation vector and an implicit activity representation vector are calculated.
[0091] In a further embodiment, based on the daily vibration consistency index, the period vibration consistency index and the vibration stability change index and integrating the motion characteristics of the vehicle, calculating the comprehensive activity representation vector and the implicit activity representation vector includes the following steps:
[0092] S41. Obtain the motion characteristics of the vehicle (referring to the characteristics of the vehicle when in motion, such as speed, acceleration, vibration of the road surface, etc. The motion characteristics can be used to describe the motion state of the vehicle), and convert the motion characteristics into an auxiliary characterization vector (the auxiliary characterization vector is a mathematical vector that contains the data of the vehicle's motion characteristics, which is used to help understand and analyze the vehicle's motion behavior).
[0093] S42, the vibration entropy vector containing the daily vibration consistency index, the period vibration consistency index and the vibration stability change index of the vibration data is fused with the auxiliary characterization vector to obtain a fused activity characterization vector; wherein entropy is an indicator to measure the randomness of data, and the vibration entropy vector is used to evaluate the stability of vehicle vibration. The fused activity characterization vector is the result of combining the auxiliary characterization vector and the vibration entropy vector. A more comprehensive representation of the vehicle motion and vibration state is obtained through fusion.
[0094] S43. After disrupting the data sequence multiple times (data processing technology, changing the time order of the data), recalculate the auxiliary representation vector and the vibration entropy vector, and fuse them again to generate a new fused activity representation vector. At the same time, input the new fused activity representation vector into the encoder-decoder model and output the implicit activity representation vector. The encoder-decoder model is used to process sequence data. The encoder part converts the input data into an internal representation, and the decoder part converts this internal representation back to the original data format or into a new output format. That is, it is used to extract the implicit activity representation vector from the fused activity representation vector, which is a deeper data representation for further analysis and problem identification. The encoder-decoder model is used to extract useful information to more accurately identify potential problems of the vehicle, which helps to improve the efficiency and effectiveness of vehicle maintenance.
[0095] S5. Use the tree structure to analyze the test data and automatically generate a test report. At the same time, input the implicit activity representation vector into the trained multi-layer perceptron (MLP), and use the trained multi-layer perceptron to make predictions and identify whether there is a loose fastener problem.
[0096] In a further embodiment, analyzing the test data using a tree structure and automatically generating a test report includes the following steps:
[0097] S51. Obtain user behavior log data, pre-process the user behavior log data and test data, and group them according to basic information such as vehicle type, test type, and test date.
[0098] S52: Build a tree structure and use the grouping information as different nodes. For example, the top-level node may be the vehicle type, the next level may be the test type, the next level may be the test date, and so on.
[0099] S53. Data access permissions are set according to the user's role and identity information; for example, test engineers can access all test data they perform, while quality management personnel may have access to all test data. User access permissions are associated with corresponding nodes in the tree structure so that users can only access data they are authorized to view.
[0100] S54. Use deep learning algorithms to analyze user behavior log data, predict user data access needs, and dynamically adjust access permissions.
[0101] S55. Integrate the test data of each node, such as aggregating all test results under the same test type, and analyze the relationship between nodes, such as comparing the test results of different vehicle types or different time periods to identify trends and patterns. That is, apply statistical methods such as mean, standard deviation, regression analysis, etc. to analyze the aggregated data to reveal the central trend, degree of variation and correlation of the data.
[0102] S56. Automatically obtain test reports based on the tree structure and analysis results. The reports include test overview, detailed results, trend analysis, problem identification and improvement suggestions.
[0103] S57. Customize the test report content based on the user's access rights to ensure that the user obtains the information they are entitled to. Users can interactively explore data through a tree structure, such as clicking on different nodes to view detailed test data and analysis results. Provide search and filter functions to help users quickly find specific test data or analysis results.
[0104] The tree structure is used to efficiently manage and analyze test data, and insightful test reports are automatically generated, thereby improving the efficiency and accuracy of decision-making. It not only improves the structuring and security of data management, but also improves the personalization and practicality of reports.
[0105] In a further embodiment, using a deep learning algorithm to analyze user behavior log data, predicting the user's data access needs, and dynamically adjusting access permissions includes the following steps:
[0106] S541. Obtain the co-occurrence frequency data of user behaviors from the user behavior log data, that is, the number of times two behaviors occur at the same time. For example, if user A and user B both click on the same product in the same period, this will be calculated as a co-occurrence event, and the behavior similarity between the users is calculated based on the co-occurrence frequency data of the user behaviors to obtain a user behavior similarity matrix, in which each element represents the co-occurrence frequency of two user behaviors.
[0107] S542, input the user behavior similarity matrix to the sparse autoencoder, extract the key feature representation of the user behavior, and reconstruct the user behavior similarity matrix from the hidden layer features, minimize the difference between the original input and the reconstructed output, and add sparsity penalty to obtain the user behavior feature representation to optimize the feature extraction and dimensionality reduction effect. The sparsity penalty is to impose a penalty on the parameters so that these parameters become sparse (that is, most of the parameters are zero).
[0108] S543: Determine the number of clusters, and determine the initial cluster center based on the principle of maximizing the user behavior modularity and combining the density and similarity of the user behavior features.
[0109] S544, iteratively optimize the clustering results until a termination condition is met or a maximum number of iterations is reached, thereby obtaining a user group division result.
[0110] S545. Calculate the user behavior modularity, evaluate the consistency within the group and the distinction between groups, and adjust the user access rights according to the user behavior modularity value to ensure that the high similarity user groups have a reasonable data access range.
[0111] Modularity is an indicator that measures the quality of group division in a network. A network refers to a structure consisting of nodes (users) and edges (relationships or interactions between users), while a group refers to a community in a network, i.e., a collection of users with similar behavior patterns or characteristics. In the context of user behavior analysis, modularity helps understand the clustering of user behavior.
[0112] Internal consistency of a group: Calculates the degree of connection between nodes within a group. A high modularity value means that the connections within the group are closer than random, indicating a high degree of consistency within the group.
[0113] Inter-group differentiation evaluation: Evaluate the differentiation between different groups, that is, there are relatively few edges between groups and relatively more edges within groups. This helps to identify user groups with similar behavior patterns.
[0114] User access rights adjustment: User access rights are adjusted based on modularity values and group structure. For example, if two users belong to the same group with highly similar behavior patterns, they will be given similar data access rights.
[0115] S546. Implement the user's access rights, monitor the user's behavior changes after the access rights are adjusted, and dynamically adjust the user's access rights according to the changes in the user's behavior.
[0116] Through this step, user access rights can be dynamically adjusted to reflect changes in user behavior and the discovery of community structure, thereby improving system security and personalized experience.
[0117] In a further embodiment, determining the number of clusters, based on the principle of maximizing the modularity of user behavior, and combining the density and similarity of user behavior features to determine the initial cluster center includes the following steps:
[0118] S5431. Obtain the user behavior modularity function, and calculate the user behavior modularity values under different numbers of clusters through an iterative process. If the user behavior modularity value reaches the maximum, the corresponding number of clusters is the optimal number of clusters.
[0119] S5432. Calculate the density (number of neighboring points) of each user behavior data point within a specific radius, sort them from high to low according to density, and select the data point with the largest density as the first cluster center.
[0120] S5433. Calculate the Euclidean distance and density between the remaining data points and the determined cluster center, pre-set the density threshold and distance threshold, and select data points that are greater than or equal to the density threshold and whose distance from the existing cluster center is greater than or equal to the distance threshold as new cluster centers.
[0121] S5434. Repeat S5432-S5433 until all initial cluster centers are determined.
[0122] Among them, the formula of user behavior modularity function is:
[0123]
[0124] In the formula, Q represents the modularity of user behavior, m represents the total number of edges in the network, A ab Represents the nodes in the adjacency matrix a andb The weight of the edge between k a and k b Respectively represent nodes a and b The degree of δ ( c a , c b ) represents the indicator function, when the node a and b If they belong to the same group, it is 1, otherwise it is 0. c a and c b Represents the nodes in the network a and b The groups you belong to.
[0125] In a further embodiment, the user behavior modularity function is obtained, and when the user behavior modularity values under different numbers of clusters are calculated through an iterative process, an optimization algorithm (such as simulated annealing, genetic algorithm, etc.) is used, with maximizing the user behavior modularity as the objective function, and the cluster center is adjusted in the iterative process until the cluster center configuration that can maximize the user behavior modularity is found.
[0126] In a further embodiment, the implicit activity representation vector is input into a trained multi-layer perceptron, and prediction is performed by the trained multi-layer perceptron to identify whether there is a fastener loosening problem, including the following steps:
[0127] The implicit activity representation vectors generated by previous vibration data are collected in advance and converted into standardized feature vectors (the format meets the requirements of MLP). The feature vectors need to be standardized, that is, the data will be scaled to a standard range, usually 0 to 1 or -1 to 1, so that the model can process it more easily.
[0128] The multi-layer perceptron is trained and evaluated using standardized feature vectors to obtain a trained multi-layer perceptron.
[0129] In a further embodiment, training and evaluating a multilayer perceptron using standardized feature vectors comprises the following steps:
[0130] Assume the structure of a multilayer perceptron, including the input layer, hidden layer (and the number of neurons), and output layer, and select the activation function (such as the sigmoid function) and the initialization weights;
[0131] Divide the standardized feature vector into training data set and test data set;
[0132] Using the gradient descent method, the multilayer perceptron is trained using the training data set, and the neuron weights are adjusted to minimize the squared error;
[0133] The performance of the multilayer perceptron is evaluated using the test data set, using indicators such as AUC-ROC curve and precision and recall.
[0134] The newly acquired implicit activity representation vector is converted into a format and fed into the trained multi-layer perceptron, which outputs a prediction result indicating whether the fastener is loose, helping the maintenance team determine whether further inspection or repair is needed.
[0135] Among them, the multi-layer perceptron (MLP) is a feedforward neural network that includes at least three layers of nodes: input layer, hidden layer, and output layer. Except for the input node, each node is a neuron with nonlinear activation function. MLP is the basis of deep learning and neural networks, and can learn complex mapping relationships between input data and output data.
[0136] S6. Based on the test report and prediction results, determine whether the fasteners are loose. If loose, record the location and degree of the loose fasteners.
[0137] Review the test report to determine if there are any signs of loose fasteners, such as unusual vibration data or changes in torque values. Review the predictions provided by the Multilayer Perceptron. Confirm if the model predicts that there is a loose fastener issue.
[0138] If both the test report and the predicted results indicate loosening, confirm the finding. If the results disagree, perform additional inspections or use other diagnostic tools to verify.
[0139] Identify and record the exact location of the loosened fastener. Evaluate the severity of the loosened fastener, including measuring the range of motion or torque loss of the loosened part.
[0140] In summary, the present invention provides a method for testing the loose chassis fasteners of the suspension and brake system. Through static testing and dynamic testing, torque values and vibration data are obtained, and a comprehensive test data set is jointly constructed for subsequent analysis. By calculating the daily vibration consistency index, period vibration consistency index and vibration stability change index of the vibration data, a quantitative evaluation of the daily operation stability of the vehicle is provided, and the necessary input data is provided for the calculation of the comprehensive activity characterization vector. The vibration data and vehicle motion characteristics are integrated to provide a comprehensive overview of vehicle activities and generate a comprehensive data representation for a multi-layer perceptron. The test data is analyzed using a tree structure to provide a structured test report, and dynamic management of data access rights is realized. A personalized test report is provided, and permission management and user behavior analysis are combined to improve data security and personalized experience. The user behavior is analyzed using a deep learning algorithm to achieve intelligent adjustment of access rights, and the calculation of user behavior modularity provides a basis for clustering and permission management. Multilayer perceptron learning is used to identify patterns of fastener loosening from historical data, and this knowledge is applied to new data to predict and diagnose potential problems. Potential fastener loosening problems are identified by multilayer perceptron, and the prediction results provide a scientific basis for maintenance decisions.
[0141] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A vehicle test method for loose chassis fasteners of suspension and brake systems, characterized in that: The whole vehicle test method for loose chassis fasteners of the suspension and brake system comprises the following steps: S1. Obtain the initial state of all fasteners; S2. Performing a static test on the vehicle and obtaining the torque values of the fasteners of the suspension system and the brake system, and performing a dynamic test on the vehicle to obtain the vibration data of the vehicle, and constructing test data based on the torque values and vibration data of the fasteners; S3. Calculate the daily vibration consistency index, period vibration consistency index and vibration stability change index of the vibration data; S4. Based on the daily vibration consistency index, the period vibration consistency index and the vibration stability change index and integrating the motion characteristics of the vehicle, a comprehensive activity representation vector and an implicit activity representation vector are calculated; S5. Analyze the test data using a tree structure, automatically generate a test report, and input the implicit activity representation vector into a trained multi-layer perceptron, and make predictions through the trained multi-layer perceptron to identify whether there is a problem of fastener looseness; S6. Based on the test report and prediction results, determine whether the fasteners are loose. If so, record the location and degree of the loose fasteners. The calculation of the daily vibration consistency index, the period vibration consistency index and the vibration stability change index of the vibration data comprises the following steps: S31, calculating the entropy value of the vibration data within one day to obtain a daily vibration consistency index, and using the daily vibration consistency index to evaluate the stability of the vehicle in daily operation; S32, by calculating the entropy value of the vibration data in different time periods, obtaining the time period vibration consistency index, and evaluating the running stability of the vehicle in each time period by the time period vibration consistency index; S33, obtaining a vibration stability change index by comparing the entropy value changes of the vibration data of the two time periods, and using the vibration stability change index to represent the change trend of the vehicle running stability; When calculating the entropy value of the vibration data within one day, the entropy value calculation formula is: ; In the formula, H Represents the entropy value of vibration data; p ( x i ) indicates the i The probability of a data point; x i Indicates i The value of the data point, n Represents the total number of data points.
2. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 1, characterized in that: The method of calculating a comprehensive activity representation vector and an implicit activity representation vector based on the daily vibration consistency index, the period vibration consistency index and the vibration stability change index and integrating the motion characteristics of the vehicle comprises the following steps: S41, obtaining motion features of the vehicle, and converting the motion features into auxiliary representation vectors; S42, fusing the vibration entropy vector including the daily vibration consistency index, the period vibration consistency index and the vibration stability change index of the vibration data with the auxiliary representation vector to obtain a fused activity representation vector; S43. After disrupting the data sequence for multiple times, recalculate the auxiliary representation vector and the vibration entropy vector, and fuse them again to generate a new fused activity representation vector. At the same time, input the new fused activity representation vector into the encoder-decoder model, and output the implicit activity representation vector.
3. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 2, characterized in that: The method of analyzing the test data using a tree structure and automatically generating a test report comprises the following steps: S51, obtaining user behavior log data, preprocessing the user behavior log data and test data, and grouping them according to vehicle type, test type and test date; S52, constructing a tree structure, and using the grouping information as different nodes; S53, setting data access permissions according to the user's role and identity information, and associating the user's access permissions with corresponding nodes in the tree structure, so that the user can only access data that he or she has permission to view; S54. Analyze user behavior log data using deep learning algorithms, predict user data access needs, and dynamically adjust access permissions; S55. Integrate the test data of each node, analyze the relationship between nodes, and identify trends and patterns; S56. Automatically obtain test reports based on tree structure and analysis results; S57. Customize the test report content according to the user's access rights to ensure that the user obtains the information he or she is entitled to access.
4. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 3, characterized in that: The method of using a deep learning algorithm to analyze user behavior log data, predict user data access requirements, and dynamically adjust access permissions includes the following steps: S541, obtaining co-occurrence frequency data of user behaviors from user behavior log data, and calculating the behavior similarity between users based on the co-occurrence frequency data of user behaviors to obtain a user behavior similarity matrix; S542, inputting the user behavior similarity matrix into the sparse autoencoder, extracting the key feature representation of the user behavior, and reconstructing the user behavior similarity matrix from the hidden layer features, and adding a sparsity penalty to obtain the user behavior feature representation; S543, determining the number of clusters, and determining the initial cluster center based on the principle of maximizing the modularity of user behavior and combining the density and similarity of user behavior features; S544, iteratively optimizing the clustering results until a termination condition is met or a maximum number of iterations is reached, thereby obtaining a user group division result; S545, calculating the user behavior modularity, evaluating the consistency within the group and the distinction between the groups, and adjusting the user access rights according to the user behavior modularity value; S546. Implement the user's access rights, monitor the user's behavior changes after the access rights are adjusted, and dynamically adjust the user's access rights according to the changes in the user's behavior.
5. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 4, characterized in that: The method of determining the number of clusters, based on the principle of maximizing the modularity of user behavior, and combining the density and similarity of user behavior features to determine the initial cluster center includes the following steps: S5431, obtaining a user behavior modularity function, and calculating the user behavior modularity values under different numbers of clusters through an iterative process. If the user behavior modularity value reaches the maximum, the corresponding number of clusters is the optimal number of clusters; S5432, calculating the density of each user behavior data point within a specific radius, and sorting them from high to low according to density, and selecting the data point with the largest density as the first cluster center; S5433, calculating the Euclidean distance and density between the remaining data points and the determined cluster center, and presetting the density threshold and distance threshold, and selecting the data points that are both greater than or equal to the density threshold and whose distance from the existing cluster center is greater than or equal to the distance threshold as the new cluster center; S5434. Repeat S5432-S5433 until all initial cluster centers are determined.
6. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 5, characterized in that: The user behavior modularity function is obtained, and when the user behavior modularity values under different numbers of clusters are calculated through an iterative process, an optimization algorithm is used with maximizing the user behavior modularity as the objective function, and the cluster center is adjusted during the iterative process until a cluster center configuration that can maximize the user behavior modularity is found.
7. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 6, characterized in that: The step of inputting the implicit activity representation vector into the trained multi-layer perceptron and performing prediction by the trained multi-layer perceptron to identify whether there is a fastener loosening problem includes the following steps: Collect implicit activity representation vectors generated by previous vibration data in advance and convert them into standardized feature vectors; Use the standardized feature vector to train and evaluate the multilayer perceptron to obtain a trained multilayer perceptron; The newly acquired implicit activity representation vector is converted into a new format and then fed into the trained multi-layer perceptron, which then outputs a prediction result indicating whether the fastener is loose.
8. A whole vehicle test method for loose chassis fasteners of a suspension and brake system according to claim 7, characterized in that: The method of training and evaluating the multilayer perceptron using the standardized feature vector comprises the following steps: Assume the structure of a multilayer perceptron, including input layer, hidden layer and output layer, and select activation function and initialization weights; Divide the standardized feature vector into training data set and test data set; Using the gradient descent method, the multilayer perceptron is trained using the training data set, and the neuron weights are adjusted to minimize the squared error; Evaluate the performance of the Multilayer Perceptron using the test dataset.
Citation Information
Patent Citations
Fastener looseness detection method and device, electronic equipment and storage medium
CN113074931A
Method for controlling consistency of vibration of surfaces of shell structures on basis of neural networks
CN104238367A
Fuzzy granulation prediction method of performance degradation of rolling bearing on the basis of information entropy
CN105590023A
Bolt connection structure loosening state recognition method based on vibrational response information
CN109506907A
Elevator fixing device loosening state detection method, computer equipment and elevator
CN111753260A
