A comprehensive automobile quality analysis method and system based on multi-dimensional indicators

Abstract

This invention discloses a comprehensive vehicle quality analysis method and system based on multidimensional indicators, relating to the field of vehicle quality analysis. The method involves collecting quality analysis data from sample vehicles, constructing a quality indicator vector after preprocessing, calculating the objective weights of the quality indicators based on the coefficient of variation of each quality indicator in the quality indicator vector, evenly distributing the total subjective weight of the quality indicators and optimizing the subjective weights, calculating the weight coefficients of each quality indicator, calculating the proximity of sample vehicles, and performing an initial grading. Based on the quality grade distribution of the sample vehicles, vehicles with quality issues are screened, and the comprehensive vehicle quality analysis results are output. This solution constructs a quality causal directed acyclic graph and improves the accuracy of locating vehicle quality issues by quantifying the intervention effect values ​​of the quality indicators.

Description

Technical Field

[0001] The present invention relates to the field of automobile quality analysis, and in particular to a comprehensive automobile quality analysis method and system based on multi-dimensional indicators. Background Art

[0002] Current automotive quality evaluation systems are primarily based on static weight assignments and fixed grading standards. Traditional methods such as the analytic hierarchy process (AHP) or entropy weighting are often used to determine indicator weights. With the advancement of multi-source, heterogeneous data acquisition technologies, IoT sensors and on-board onboard diagnostic (OBD) systems can now acquire quantitative indicators for dimensions such as power, safety, and environmental protection in real time, laying the foundation for comprehensive multi-dimensional evaluation. Existing research typically uses the Top-of-Search by Approximately Ideal Solutions (TOPSIS) method combined with fuzzy C-means clustering to grade vehicles. However, this ideal solution distance metric relies on Euclidean distance, which can easily lead to ranking ambiguity in vertically bisected regions. Furthermore, weight calculations often focus on objective statistical features (such as the coefficient of variation) and lack adaptive optimization mechanisms that account for the correlation between user preferences and dynamic quality.

[0003] Existing technologies have significant limitations: First, static weight allocation strategies cannot adapt to the differentiated quality requirements of multiple vehicle models. For example, the dimensional differences in environmental performance indicators between fuel-powered and electric vehicles are not dynamically calibrated, resulting in reduced comparability of cross-vehicle evaluations. Second, the TOPSIS method based on Euclidean distance is prone to classification failure at symmetrical locations of the ideal solution. When the ratio of the distance between a sample vehicle and the positive and negative ideal solutions is equal, it cannot effectively distinguish quality levels, resulting in ambiguous evaluation results. Furthermore, subjective weights are often initialized using an equal-divide method, without a feedback optimization mechanism based on the vehicle's actual quality performance (such as C-NCAP scores), which reduces the engineering guidance value of the evaluation system. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a comprehensive automobile quality analysis method based on multi-dimensional indicators to solve the problem that the static weight allocation in the traditional automobile quality evaluation system is difficult to adapt to the dynamic evaluation needs of multiple models, and the grading method based on Euclidean distance has evaluation ambiguity.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a comprehensive automobile quality analysis method based on multidimensional indicators, comprising collecting quality analysis data of sample automobiles, performing preprocessing, and constructing a quality indicator vector; the quality analysis data includes the sample automobile type, the sample automobile C-NCAP score, and the original quality data;

[0008] By calculating the coefficient of variation of each quality indicator in the quality indicator vector, the objective weight of each quality indicator is generated. By evenly distributing the total subjective weight of the quality indicators, the subjective weight of each quality indicator is obtained. The subjective weight is then optimized. Based on the objective weight and the subjective weight, the weight coefficient of each quality indicator is calculated.

[0009] Based on the quality indicator vector, a decision matrix is ​​constructed, and the decision matrix is ​​weighted according to the weight coefficient of each quality indicator to obtain a weighted decision matrix. According to the weighted decision matrix, positive and negative ideal solutions are set, and the quality of each sample car is initially graded by calculating the proximity of each sample car. The initial grading result is verified to generate the quality grade distribution weight coefficient of the sample car;

[0010] Based on the quality grade distribution of sample cars, vehicles with quality problems are screened. Based on a directed acyclic graph, with the quality indicator vector as the cause node, the comprehensive quality coefficient as the result node, and the quality indicators pointing to the comprehensive quality coefficient as the edges, a quality causal graph for vehicles with quality problems is constructed. By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, the problematic quality indicators of vehicles with quality problems are identified and output as the comprehensive analysis result of automobile quality.

[0011] As a preferred solution of the comprehensive automobile quality analysis method based on multi-dimensional indicators of the present invention, the objective weights of each quality indicator are generated in the following specific steps:

[0012] Based on the mean and standard deviation of each quality indicator in the quality indicator vector in all sample cars, the coefficient of variation method is used to calculate the coefficient of variation of each quality indicator;

[0013] The coefficient of variation of each quality indicator is normalized to obtain the objective weight of each quality indicator.

[0014] As a preferred solution of the comprehensive automobile quality analysis method based on multi-dimensional indicators of the present invention, the optimization of the subjective weight is carried out in the following specific steps:

[0015] The state space is defined as the mean and standard deviation of each quality indicator in all sample cars, and the action space is the increase or decrease of the subjective weight;

[0016] The subjective quality coefficient is calculated based on the product of the quality index of each sample car and the corresponding subjective weight. The quality matching rate is obtained by calculating the Pearson correlation coefficient between the subjective quality coefficient and the C-NCAP score.

[0017] Define a reward function based on the quality matching rate and subjective quality coefficient of the sample cars;

[0018] Based on the state space, action space, subjective weight and reward function value, a deep Q network is used to optimize the subjective weight.

[0019] As a preferred solution of the comprehensive automobile quality analysis method based on multi-dimensional indicators of the present invention, the specific steps of calculating the proximity of each sample car are as follows:

[0020] Constructing a decision matrix based on the quality indicator vector and the number of sample cars, wherein each row of the decision matrix represents a sample car and each column represents a quality indicator;

[0021] The quality index of each column of the decision matrix is ​​weighted using the weight coefficient of the quality index to generate a weighted decision matrix;

[0022] Setting a positive ideal solution and a negative ideal solution; the positive ideal solution is the maximum value of each column of the weighted decision matrix, and the negative ideal solution is the minimum value of each column of the weighted decision matrix, and generating a positive ideal solution vector and a negative ideal solution vector;

[0023] Calculate the relative entropy of each sample car with the positive ideal solution vector and the negative ideal solution vector;

[0024] Based on the relative entropy of the positive and negative ideal solution vectors, the proximity of each sample car is calculated.

[0025] As a preferred solution of the comprehensive automobile quality analysis method based on multi-dimensional indicators of the present invention, the specific steps of verifying the initial classification result are as follows:

[0026] Using C-NCAP scores as an external validation standard, define consistency rules between the C-NCAP scores of sample vehicles and the initial rating results;

[0027] Based on the indicator function, the consistency rate between the C-NCAP score and the initial classification result was calculated;

[0028] A consistency rate threshold is set based on the statistical distribution of the C-NCAP scores and initial grading results of all sample vehicles;

[0029] When the consistency rate is greater than the consistency rate threshold, it means that the initial grading result is valid and the quality grade distribution of the sample car is generated. Otherwise, the initial grading result is regenerated until the consistency rate is greater than the consistency rate threshold.

[0030] As a preferred solution of the comprehensive automobile quality analysis method based on multi-dimensional indicators of the present invention, the specific steps of identifying the problem quality indicators of vehicles with quality problems are as follows:

[0031] Based on the quality cause-effect diagram, calculate the intervention effect value of each quality indicator on the quality level;

[0032] For vehicles with quality problems, the quality indicator with the largest intervention effect value is used as the problem quality indicator.

[0033] As a preferred solution of the comprehensive automobile quality analysis method based on multi-dimensional indicators described in the present invention, the calculation of the weight coefficient of each quality indicator refers to weighted summation of the objective weight and subjective weight of each quality indicator.

[0034] In a second aspect, the present invention provides a comprehensive automobile quality analysis system based on multi-dimensional indicators, comprising:

[0035] A quality indicator module is used to collect quality analysis data of sample vehicles and construct a quality indicator vector after preprocessing; the quality analysis data includes the sample vehicle type, the sample vehicle C-NCAP score, and the original quality data;

[0036] The weight coefficient module is used to generate the objective weight of each quality indicator by calculating the coefficient of variation of each quality indicator in the quality indicator vector, obtain the subjective weight of each quality indicator by evenly distributing the total subjective weight of the quality indicators, optimize the subjective weight, and calculate the weight coefficient of each quality indicator based on the objective weight and subjective weight;

[0037] a quality grading module, configured to construct a decision matrix based on the quality indicator vector, weight the decision matrix according to the weight coefficient of each quality indicator to obtain a weighted decision matrix, set positive and negative ideal solutions based on the weighted decision matrix, perform an initial grading of the quality of the sample cars by calculating the proximity of each sample car, verify the initial grading results, and generate a quality grade distribution of the sample cars;

[0038] The analysis result module is used to screen vehicles with quality problems based on the quality grade distribution of sample vehicles. Based on a directed acyclic graph, the quality indicator vector is used as the cause node, the comprehensive quality coefficient is used as the result node, and each quality indicator points to the comprehensive quality coefficient as an edge. A quality causal graph of vehicles with quality problems is constructed. By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, the problem quality indicators of vehicles with quality problems are identified and output as the comprehensive analysis result of the vehicle quality.

[0039] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the comprehensive automobile quality analysis method based on multidimensional indicators as described in the first aspect of the present invention is implemented.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the comprehensive automobile quality analysis method based on multidimensional indicators as described in the first aspect of the present invention.

[0041] The beneficial effects of the present invention are as follows: the present invention designs a differentiated standard processing flow to address the problem of dimensional incompatibility of fuel / pure electric vehicle indicators, adopts relative entropy instead of Euclidean distance to calculate the proximity between samples and ideal solutions, avoids the sorting failure problem of traditional methods in the vertical domain of the hyperplane, and combines the grade verification mechanism of fuzzy C-means clustering to improve the grading accuracy of unqualified cars, constructs a quality causal directed acyclic graph, and improves the positioning accuracy of automobile quality problems by quantifying the intervention effect value of quality indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 any creative work.

[0043] Figure 1 This is a flowchart of the comprehensive analysis method of automobile quality based on multi-dimensional indicators.

[0044] Figure 2 A flow chart is generated for the weight coefficients of the comprehensive automobile quality analysis method based on multi-dimensional indicators.

[0045] Figure 3 This is a flowchart for the initial classification and verification of the comprehensive analysis method for automobile quality based on multi-dimensional indicators.

[0046] Figure 4 This is a flowchart for problem indicator identification of comprehensive automobile quality analysis method based on multi-dimensional indicators. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0050] Reference Figures 1 to 4, is an embodiment of the present invention, which provides a comprehensive automobile quality analysis method based on multi-dimensional indicators, comprising the following steps:

[0051] S1. Collect the quality analysis data of the sample car, perform preprocessing and construct the quality index vector.

[0052] Obtain the quality analysis data of the sample car, including the sample car type, sample car C-NCAP score and original quality data. The specific contents are as follows:

[0053] Specific information about the sample vehicle type, including whether it is a fuel / hybrid vehicle or a new energy pure electric vehicle, as well as the sample vehicle's C-NCAP score, is determined from the manufacturer's database using the vehicle identification number (VIN). Sample vehicles of each type are randomly selected from the target production batch, and raw quality data for each sample vehicle is collected in real time using IoT technology and dedicated equipment. The C-NCAP score for the sample vehicle is China's authoritative automotive safety rating system, operated by the China Automotive Testing and Assessment Management Center. Its score comprehensively evaluates a vehicle's performance in three major areas: occupant protection, pedestrian protection (VRU protection), and active safety. This score is common knowledge in the field and is not elaborated on in detail.

[0054] For each raw quality data of each sample car, calculate the mean and standard deviation of the raw quality data of all sample cars and standardize them to , in order to eliminate the dimensional differences between different indicators, we can obtain the following indicators including power index, economic index, environmental index, safety index, comfort index, user experience index and manufacturing consistency index:

[0055] Power Index: Using an acceleration test device, measure the time required for each sample vehicle to accelerate from 0 to 100 km / h, as well as the maximum speed of each device. The calculation formula is as follows:

[0056] ;

[0057] in, is the power indicator, is the average maximum speed of all sample cars, is the standard deviation of the maximum speed of all sample cars, For the The maximum speed of the sample cars, is the average 0-100km / h acceleration time of all sample cars, For the 0-100km / h acceleration time of sample cars, is the standard deviation of the 0-100km / h acceleration time, is the index variable for the number of sample cars;

[0058] Economic indicators: For fuel / hybrid vehicles, the lowest fuel consumption per 100 kilometers is measured under NEDC operating conditions through the vehicle's OBD interface. For pure electric vehicles, the lowest power consumption per 100 kilometers is measured under CLTC operating conditions through the OBD interface. The calculation formula is as follows:

[0059] ;

[0060] in, is an economic indicator, For the The lowest fuel consumption per 100 kilometers (fuel / hybrid vehicles) or the lowest power consumption per 100 kilometers (pure electric vehicles) of the sample vehicles, is the average of the lowest fuel consumption or lowest power consumption per 100 kilometers of all sample cars, is the standard deviation of the lowest fuel consumption or lowest electricity consumption per 100 kilometers for all sample vehicles;

[0061] Environmental indicators: For fuel / hybrid vehicles, exhaust emissions are measured using emission test equipment. For pure electric vehicles, the pure electric range is measured using endurance test equipment under CLTC conditions. The calculation formula is as follows:

[0062] For gasoline / hybrid vehicles:

[0063] ;

[0064] For pure electric vehicles:

[0065] ;

[0066] in, It is an environmental performance index for fuel / hybrid vehicles. For the The exhaust emissions of a sample of fuel / hybrid vehicles, is the average exhaust emission of all fuel / hybrid vehicles, is the standard deviation of exhaust emissions of all fuel / hybrid vehicles, It is the environmental protection index of pure electric vehicles. For the The pure electric range of the sample cars is pure electric. is the average of the pure electric range of all pure electric vehicles, is the standard deviation of the pure electric range of all pure electric vehicles;

[0067] Safety indicators: The frontal collision energy absorption rate is measured by the collision test sensor, and the active safety unit response time is measured by the simulated pedestrian detection test equipment. The calculation formula is as follows:

[0068] ;

[0069] in, is a safety indicator, For the The frontal collision energy absorption rate of the sample cars, is the mean of the frontal collision energy absorption rate of all sample cars, is the standard deviation of the frontal collision energy absorption rate of all sample cars, For the The active safety unit response time of the sample vehicles, is the mean response time of all sample vehicles’ active safety units, is the standard deviation of the response time of the active safety units of all sample vehicles;

[0070] Comfort index: The interior noise level is measured using a sound level meter at a constant speed of 60 km / h. The suspension vibration attenuation rate is measured using a vibration test bench. The calculation formula is as follows:

[0071] ;

[0072] in, is the comfort index, and Respectively The interior noise level and suspension vibration attenuation rate of the sample cars, and are the mean values ​​of interior noise level and suspension vibration attenuation rate of all sample cars, and are the standard deviations of the interior noise level and suspension vibration attenuation rate of all sample cars, respectively;

[0073] User experience indicators: User satisfaction scores (out of 100) are collected through the online questionnaire API, and human-computer interaction response time is extracted from the vehicle system log. The calculation formula is as follows:

[0074] ;

[0075] in, For user experience indicators, and Respectively User satisfaction scores and human-computer interaction response time of sample cars, and are the means of all sample car user satisfaction scores and human-computer interaction response time, and are the standard deviations of all sample car user satisfaction scores and human-computer interaction response time;

[0076] Manufacturing consistency index: The dimensional deviation rate and process parameter fluctuation rate of key components (such as wheels, anti-collision beam A / B pillars, etc.) are measured using a high-precision 3D scanner. The calculation formula is as follows:

[0077] ;

[0078] in, To create consistency indicators, and Respectively The dimensional deviation rate and process parameter fluctuation rate of key components of sample automobiles, and are the mean of the dimensional deviation rate and process parameter fluctuation rate of all sample automotive key components, and are the standard deviations of the dimensional deviation rate and process parameter fluctuation rate of all sample automotive key components respectively;

[0079] Use the Z-score method to detect outliers in each quality indicator of each sample car, and replace the detected outliers with the mean value of the quality indicator;

[0080] For indicators with missing values, the K-nearest neighbor interpolation method is used to fill in the missing values. Specifically, if a certain indicator of a sample car is missing, the five sample cars with the closest Euclidean distance in other non-missing indicator values ​​are selected, and the mean value of the missing indicator of these five sample cars is calculated as the interpolation result.

[0081] The quality index of each sample car is organized into a vector form and the quality index vector is output.

[0082] S2. Generate the objective weight of each quality indicator by calculating the coefficient of variation of each quality indicator in the quality indicator vector. Obtain the subjective weight of each quality indicator by evenly distributing the total subjective weight of the quality indicators. Optimize the subjective weights and calculate the weight coefficient of each quality indicator based on the objective weights and subjective weights.

[0083] Based on the mean and standard deviation of each quality indicator in the quality indicator vector in all sample cars, the coefficient of variation method is used to calculate the coefficient of variation of each quality indicator. The calculation formula is as follows:

[0084] ;

[0085] in, For the The coefficient of variation of the quality index, is the quantity index of the quality indicator, is the number of sample cars, Indicates the The first quality indicators;

[0086] Normalize the coefficient of variation of each quality indicator to obtain the objective weight of each quality indicator;

[0087] Evenly distribute the total subjective weight, that is, divide the total subjective weight equally to initialize the subjective weight of each quality indicator. The total subjective weight is the sum of the subjective weights of each quality indicator and is 1.

[0088] The deep Q network is used to optimize the subjective weights. The specific steps are as follows:

[0089] The state space is defined as the mean and standard deviation of each quality indicator in all sample cars, and the action space is the increase or decrease of the subjective weight;

[0090] The subjective quality coefficient is calculated based on the product of the quality index of each sample car and the corresponding subjective weight. The quality matching rate is obtained by calculating the Pearson correlation coefficient between the subjective quality coefficient and the C-NCAP score.

[0091] Based on the quality matching rate and subjective quality coefficient of the sample car, the reward function is defined and calculated as follows:

[0092] ;

[0093] ;

[0094] in, For the The subjective quality coefficient of the sample cars, is the number of quality indicators, is the quality index and the corresponding subjective weight, the initial value is , is the reward function value, is the quality matching rate, is the mean of the subjective quality coefficients of all sample cars, and They are and The weight coefficient of , ;

[0095] Based on the deep Q-network state space, action space, subjective weights, and reward function values, the deep Q-network is trained to optimize the subjective weights of each quality indicator. The specific steps are as follows:

[0096] Among them, the deep Q network includes an input layer, three fully connected layers and an output layer. It takes the state space as input and outputs the state value function to obtain the optimal action space. In each iteration, the action space should be selected using the ϵ-greedy strategy based on the current state space. The subjective weight and reward function value are updated according to the selected action space. Based on the mean square error, the loss function of the deep Q network is set. By minimizing the loss function, the network parameters of the deep Q network are updated. When the deep Q network reaches the convergence condition, such as the change in the reward function value is less than 0.01, the training is stopped and the optimal subjective weight is output. The above deep Q network training process is existing technology and will not be described in detail.

[0097] The weight coefficient of each quality indicator is obtained by weighted summing the objective weight and subjective weight of each quality indicator.

[0098] S3. Based on the quality indicator vector, a decision matrix is ​​constructed, and the decision matrix is ​​weighted according to the weight coefficient of each quality indicator to obtain a weighted decision matrix. According to the weighted decision matrix, positive and negative ideal solutions are set, and the quality of each sample car is initially graded by calculating the proximity of each sample car. The initial grading results are verified to generate a quality grade distribution of the sample cars.

[0099] Calculate the proximity of each sample car. The specific steps are as follows:

[0100] Based on the quality index vector and the number of sample cars, a decision matrix is ​​constructed, which can be expressed as Each row of the decision matrix represents a sample car, and each column represents a quality indicator. The "7" in the expression of the decision matrix means that the number of quality indicators is 7. See step S1 for details. The quality indicators in each column of the decision matrix are weighted using the weight coefficients of each quality indicator to generate a weighted decision matrix.

[0101] Set positive and negative ideal solutions. Since the quality indicators in the quality indicator vector are all benefit-oriented indicators (the larger the value, the better), the positive ideal solution is the maximum value of each column of the weighted decision matrix, and the negative ideal solution is the minimum value of each column of the weighted decision matrix. Generate a positive ideal solution vector and a negative ideal solution vector;

[0102] Calculate the relative entropy of each sample car and the positive and negative ideal solution vectors using the following formula:

[0103] ;

[0104] in, Indicates the The relative entropy between the sample cars and the positive ideal solution vector, is the first positive ideal solution vector Levi values, Represents the weighted decision matrix The first Levi values;

[0105] The relative entropy of each sample car and the negative ideal solution vector The calculation of is the same as the positive ideal solution vector;

[0106] Based on the relative entropy of the positive ideal solution and the relative entropy of the negative ideal solution, the proximity of each sample car is calculated. The proximity indicates the relative closeness of the sample car to the positive ideal solution. The closer the value is to 1, the higher the quality. The calculation formula is as follows:

[0107] ;

[0108] in, For the The proximity of the sample cars ranges from ;

[0109] It should be noted that the traditional approach to ideal solution ranking method uses Euclidean distance to measure the closeness of a sample to the ideal solution, which leads to ranking ambiguity at the perpendicular bisector of the ideal solution (i.e., the distances between two samples and the positive and negative ideal solutions are equal, making it impossible to distinguish between the good and the bad). This embodiment uses the relative entropy of the positive and negative ideal solution vectors instead of the Euclidean distance in the traditional approach to ideal solution ranking method, solving this problem by measuring the difference in information distribution and providing a more accurate ranking.

[0110] Based on the proximity value, fuzzy C-means clustering is used to perform initial classification of sample cars and generate initial classification results. The specific steps are as follows:

[0111] Set the sample car quality grade categories, such as excellent, qualified, and unqualified, and initialize the sample car quality membership matrix. Each row represents a sample car, and each column represents the sample car's membership. The sample car quality membership represents the probability of the sample car belonging to a certain grade. The initial value is obtained by random assignment.

[0112] The objective function is set based on minimizing the sum of the squares of the weighted Euclidean distances between the proximity and the cluster center. The initial value of the cluster center can be set to the quartile of the proximity. The membership of the sample car mass and the cluster center are iteratively updated. Specifically, the membership of the sample car mass is updated according to the inverse of the Euclidean distance between the proximity and the cluster center, and the cluster center is updated by the mean of the proximity weighted by the membership.

[0113] When the objective function reaches the convergence condition, for example, the objective function change is less than 0.001 or the maximum number of iterations is reached, the initial classification result of each sample car is obtained according to the maximum membership degree of each sample car;

[0114] Verify the initial classification results of the sample car. The specific steps are as follows:

[0115] The C-NCAP score is used as an external verification standard, and the consistency rules between the C-NCAP score of the sample car and the initial classification result are defined. For example, a C-NCAP score of 80 is excellent, and a C-NCAP score of If the C-NCAP score is within the range, it is qualified, and if the C-NCAP score is less than 60, it is unqualified. Based on the C-NCAP score and the initial grading result of each sample vehicle, the consistency rate of the C-NCAP score and the initial grading result is calculated. Specifically, the indicator function is used to compare the C-NCAP score of each sample vehicle with the initial grading result. When the initial grading result of the sample vehicle is consistent with the C-NCAP score, for example, the initial grading result of a sample vehicle is qualified and the C-NCAP score is 78, it is consistent. Otherwise, it is inconsistent. The ratio of sample vehicles whose initial grading results are consistent with the C-NCAP score to the total number of sample vehicles is calculated to obtain the consistency rate;

[0116] According to the statistical distribution of the C-NCAP scores of all sample vehicles and the initial classification results, a consistency rate threshold is set, for example, 0.95;

[0117] When the consistency rate is greater than the consistency rate threshold, it means that the initial grading result is valid and the quality grade distribution of the sample cars is generated. Otherwise, the initial cluster center of the fuzzy C-means clustering is adjusted and the initial grading result is regenerated until the consistency rate is greater than the consistency rate threshold.

[0118] S4. Based on the quality grade distribution of sample cars, screen out vehicles with quality problems. Based on a directed acyclic graph, with the quality indicator vector as the cause node, the comprehensive quality coefficient as the result node, and each quality indicator pointing to the comprehensive quality coefficient as the edge, construct a quality causal graph for vehicles with quality problems. By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, identify the problem quality indicators of vehicles with quality problems and output them as the comprehensive analysis result of automobile quality.

[0119] Based on the quality grade distribution of sample cars, sample cars with unqualified grades are screened out to obtain vehicles with quality problems;

[0120] Construct a quality causal graph for vehicles with quality issues. Specifically, the quality indicators of vehicles with quality issues are used as cause nodes, and the quality levels are used as result nodes. Edges are directed from each indicator to the quality level, reflecting the impact of the quality indicators on quality. To simplify the quality causal graph, it is assumed that there is no direct causal relationship between the quality indicators.

[0121] By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, the problem quality indicators of vehicles with quality problems are identified. The specific steps are as follows:

[0122] Based on the quality cause-and-effect diagram, the Do-Calculus method is used to calculate the intervention effect value of each quality indicator on the quality grade. Specifically, for each vehicle with quality problems, the probability that the quality grade of the vehicle with quality problems will still be unqualified when a certain quality indicator is intervened to the ideal value (such as the maximum standardized value, 1) is evaluated. The intervention effect value quantifies the probability that the vehicle quality will still not meet the standard after the quality indicator is improved. The higher the probability, the greater the contribution of the indicator to the quality problem.

[0123] For vehicles with quality problems, the quality indicator with the largest intervention effect value is used as the problem quality indicator.

[0124] This embodiment also provides a comprehensive automobile quality analysis system based on multi-dimensional indicators, including:

[0125] The quality index module is used to collect the quality analysis data of the sample car and construct the quality index vector after preprocessing; the quality analysis data includes the sample car type, the sample car C-NCAP score and the original quality data; the weight coefficient module is used to

[0126] By calculating the coefficient of variation of each quality indicator in the quality indicator vector, the objective weight of each quality indicator is generated. By evenly distributing the total subjective weight of the quality indicator, the subjective weight of each quality indicator is obtained, and the subjective weight is optimized. Based on the objective weight and subjective weight, the weight coefficient of each quality indicator is calculated; the quality grade module is used

[0127] Based on the quality indicator vector, a decision matrix is ​​constructed, and the decision matrix is ​​weighted according to the weight coefficient of each quality indicator to obtain a weighted decision matrix. According to the weighted decision matrix, positive and negative ideal solutions are set, and the quality of the sample cars is initially graded by calculating the proximity of each sample car. The initial grading results are verified to generate a quality grade distribution of the sample cars; an analysis result module is used to screen vehicles with quality problems based on the quality grade distribution of the sample cars. Based on a directed acyclic graph, the quality indicator vector is used as the cause node, the comprehensive quality coefficient is used as the result node, and each quality indicator points to the comprehensive quality coefficient as an edge. A quality causal graph of vehicles with quality problems is constructed. By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, the problem quality indicators of vehicles with quality problems are identified and output as the comprehensive analysis result of the car quality.

[0128] This embodiment also provides a computer device suitable for the case of a comprehensive automobile quality analysis method based on multidimensional indicators, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the comprehensive automobile quality analysis method based on multidimensional indicators proposed in the above embodiment.

[0129] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0130] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the comprehensive automobile quality analysis method based on multi-dimensional indicators as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0131] In summary, the present invention designs a differentiated standard processing flow to address the dimensional incompatibility problem of fuel / pure electric vehicle indicators, adopts relative entropy instead of Euclidean distance to calculate the proximity between samples and ideal solutions, avoids the sorting failure problem of traditional methods in the vertical domain of the hyperplane, and combines the grade verification mechanism of fuzzy C-means clustering to improve the grading accuracy of unqualified vehicles, constructs a quality causal directed acyclic graph, and improves the positioning accuracy of automobile quality problems by quantifying the intervention effect value of quality indicators.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A comprehensive automobile quality analysis method based on multi-dimensional indicators, characterized by: include, Collecting quality analysis data of sample vehicles and constructing a quality index vector after preprocessing; the quality analysis data includes the sample vehicle type, the sample vehicle C-NCAP score, and the original quality data; By calculating the coefficient of variation of each quality indicator in the quality indicator vector, the objective weight of each quality indicator is generated. By evenly distributing the total subjective weight of the quality indicators, the subjective weight of each quality indicator is obtained. The subjective weight is then optimized. Based on the objective weight and the subjective weight, the weight coefficient of each quality indicator is calculated. Based on the quality indicator vector, a decision matrix is ​​constructed, and the decision matrix is ​​weighted according to the weight coefficient of each quality indicator to obtain a weighted decision matrix. According to the weighted decision matrix, positive and negative ideal solutions are set, and the quality of each sample car is initially graded by calculating the proximity of each sample car. The initial grading result is verified to generate a quality grade distribution of the sample cars. Based on the quality grade distribution of sample cars, vehicles with quality problems are screened. Based on a directed acyclic graph, with the quality indicator vector as the cause node, the comprehensive quality coefficient as the result node, and the quality indicators pointing to the comprehensive quality coefficient as the edges, a quality causal graph for vehicles with quality problems is constructed. By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, the problematic quality indicators of vehicles with quality problems are identified and output as the comprehensive analysis result of automobile quality.

2. The comprehensive automobile quality analysis method based on multi-dimensional indicators according to claim 1, characterized in that: The specific steps for generating the objective weight of each quality indicator are as follows: Based on the mean and standard deviation of each quality indicator in the quality indicator vector in all sample cars, the coefficient of variation method is used to calculate the coefficient of variation of each quality indicator; The coefficient of variation of each quality indicator is normalized to obtain the objective weight of each quality indicator.

3. The comprehensive automobile quality analysis method based on multi-dimensional indicators according to claim 1, characterized in that: The specific steps for optimizing the subjective weight are as follows: The state space is defined as the mean and standard deviation of each quality indicator in all sample cars, and the action space is the increase or decrease of the subjective weight; The subjective quality coefficient is calculated based on the product of the quality index of each sample car and the corresponding subjective weight. The quality matching rate is obtained by calculating the Pearson correlation coefficient between the subjective quality coefficient and the C-NCAP score. Define a reward function based on the quality matching rate and subjective quality coefficient of the sample cars; Based on the state space, action space, subjective weight and reward function value, a deep Q network is used to optimize the subjective weight.

4. The comprehensive automobile quality analysis method based on multi-dimensional indicators according to claim 1, characterized in that: The specific steps for calculating the proximity of each sample car are as follows: Constructing a decision matrix based on the quality indicator vector and the number of sample cars, wherein each row of the decision matrix represents a sample car and each column represents a quality indicator; The quality index of each column of the decision matrix is ​​weighted using the weight coefficient of the quality index to generate a weighted decision matrix; Setting a positive ideal solution and a negative ideal solution; the positive ideal solution is the maximum value of each column of the weighted decision matrix, and the negative ideal solution is the minimum value of each column of the weighted decision matrix, and generating a positive ideal solution vector and a negative ideal solution vector; Calculate the relative entropy of each sample car with the positive ideal solution vector and the negative ideal solution vector; Based on the relative entropy of the positive and negative ideal solution vectors, the proximity of each sample car is calculated.

5. The comprehensive automobile quality analysis method based on multi-dimensional indicators according to claim 1 is characterized in that: The specific steps for verifying the initial classification results are as follows: Using C-NCAP scores as an external validation standard, define consistency rules between the C-NCAP scores of sample vehicles and the initial rating results; Based on the indicator function, the consistency rate between the C-NCAP score and the initial classification result was calculated; A consistency rate threshold is set based on the statistical distribution of the C-NCAP scores and initial grading results of all sample vehicles; When the consistency rate is greater than the consistency rate threshold, it means that the initial grading result is valid and the quality grade distribution of the sample car is generated. Otherwise, the initial grading result is regenerated until the consistency rate is greater than the consistency rate threshold.

6. The comprehensive automobile quality analysis method based on multi-dimensional indicators according to claim 1, characterized in that: The specific steps for identifying the problem quality indicators of vehicles with quality problems are as follows: Based on the quality cause-effect diagram, calculate the intervention effect value of each quality indicator on the quality level; For vehicles with quality problems, the quality indicator with the largest intervention effect value is used as the problem quality indicator.

7. The comprehensive automobile quality analysis method based on multi-dimensional indicators according to claim 1 is characterized in that: The calculation of the weight coefficient of each quality indicator refers to performing a weighted summation of the objective weight and subjective weight of each quality indicator.

8. A comprehensive automobile quality analysis system based on multidimensional indicators, based on the comprehensive automobile quality analysis method based on multidimensional indicators according to any one of claims 1 to 7, characterized in that: include, A quality indicator module is used to collect quality analysis data of sample vehicles and construct a quality indicator vector after preprocessing; the quality analysis data includes the sample vehicle type, the sample vehicle C-NCAP score, and the original quality data; The weight coefficient module is used to generate the objective weight of each quality indicator by calculating the coefficient of variation of each quality indicator in the quality indicator vector, obtain the subjective weight of each quality indicator by evenly distributing the total subjective weight of the quality indicators, optimize the subjective weight, and calculate the weight coefficient of each quality indicator based on the objective weight and subjective weight; a quality grading module, configured to construct a decision matrix based on the quality indicator vector, weight the decision matrix according to the weight coefficient of each quality indicator to obtain a weighted decision matrix, set positive and negative ideal solutions based on the weighted decision matrix, perform an initial grading of the quality of the sample cars by calculating the proximity of each sample car, verify the initial grading results, and generate a quality grade distribution of the sample cars; The analysis result module is used to screen vehicles with quality problems based on the quality grade distribution of sample vehicles. Based on a directed acyclic graph, the quality indicator vector is used as the cause node, the comprehensive quality coefficient is used as the result node, and each quality indicator points to the comprehensive quality coefficient as an edge. A quality causal graph of vehicles with quality problems is constructed. By calculating the intervention effect value of each quality indicator on the comprehensive quality coefficient, the problem quality indicators of vehicles with quality problems are identified and output as the comprehensive analysis result of the vehicle quality.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of the comprehensive automobile quality analysis method based on multidimensional indicators described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the comprehensive automobile quality analysis method based on multidimensional indicators described in any one of claims 1 to 7 are implemented.

Images