Air spring safety risk detection method and system

Abstract

The embodiment of the application provides a kind of air spring safety risk detection method and system, comprising: obtaining the monitoring data of multiple sensors in air spring;The monitoring data vector set is divided into multiple subsets, to obtain the discretization monitoring data set;The rule parameter corresponding to the discretization monitoring data set is generated, to obtain the sorted rule parameter, to obtain the optimized weighted association classifier;New monitoring data is converted into discretization monitoring data set, and is classified by the optimized weighted association classifier, to obtain the risk level corresponding to air spring, by deploying multiple sensors, the comprehensive perception of the multidimensional health state of air spring is realized, the defects that only monitor single or a small number of parameters in the prior art are overcome, the classification accuracy of the association classifier is significantly improved, the classifier parameters are continuously optimized by multi-strategy improved grey wolf algorithm, so that the system can adapt to different working conditions and environmental changes, with strong generalization ability and robustness.

Description

Technical Field

[0001] This invention relates to the field of air spring testing technology, and in particular to an air spring safety risk testing method, an air spring safety risk testing system, a computer device, and a storage medium. Background Technology

[0002] As a key elastic component in vehicle suspension and rail transit systems, the working condition of air springs directly affects the overall vehicle's driving safety, ride comfort, and stability. During use, air springs can experience performance degradation or even sudden failure due to factors such as prolonged exposure to alternating loads, rubber material aging, and gas leakage. In severe cases, this can lead to loss of vehicle control and serious safety accidents. Therefore, effective safety risk detection of air springs is of significant practical importance.

[0003] Currently, existing air spring safety risk detection technologies have several shortcomings. Most existing systems rely on fixed thresholds for fault diagnosis, remaining at a passive "fault-based alarm" stage and failing to achieve early fault identification and predictive maintenance. While some systems have introduced rule-based or threshold-based judgment logic, they still cannot learn fault patterns from historical data. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide an air spring safety risk detection method, an air spring safety risk detection system, a computer device, and a storage medium to overcome or at least partially solve the above problems.

[0005] To address the aforementioned problems, this invention discloses a method for detecting safety risks of air springs, comprising: Acquire monitoring data from multiple sensors in the air spring, and convert the monitoring data into a monitoring data vector set; The monitoring data vector set is divided into multiple subsets. A Gaussian mixture model is fitted to each subset for discretization. The optimal number of Gaussian components is determined, multiple discrete intervals are generated, and an interval identifier is assigned to each discrete interval to obtain the discretized monitoring dataset. Generate rule parameters corresponding to the discrete monitoring dataset, and sort the rule parameters using a weighted association classifier to obtain the sorted rule parameters; The weighted association classifier is optimized by using a multi-strategy gray wolf population model to obtain an optimized weighted association classifier. The new monitoring data is converted into a discrete monitoring dataset, which is then classified using the optimized weighted association classifier to obtain the risk level corresponding to the air spring.

[0006] Preferably, the step of acquiring monitoring data from multiple sensors in the air spring and converting the monitoring data into a monitoring data vector set includes: Collect four-dimensional monitoring data corresponding to the air pressure sensor, displacement sensor, temperature sensor, and acceleration sensor in the air spring; The four-dimensional monitoring data is preprocessed by normalization to obtain a monitoring data vector set.

[0007] Preferably, the monitoring data vector set is divided into multiple subsets, and a Gaussian mixture model is fitted to each subset for discretization to determine the optimal number of Gaussian components, generating multiple discrete intervals. An interval identifier is assigned to each discrete interval to obtain a discretized monitoring dataset, including: Obtain the state category data corresponding to the subset; For each subset, the expectation-maximization algorithm is used to fit the Gaussian mixture model to obtain the probability density function of the Gaussian mixture model. The probability density function of the Gaussian mixture model is then iteratively updated to determine the optimal number of Gaussian components. Calculate the mean and standard deviation of each Gaussian component to generate initial intervals. Use the analogy ratio to merge or delete overlapping initial intervals to generate multiple discrete intervals. Assign an interval identifier to each discrete interval to obtain a discretized monitoring dataset.

[0008] Preferably, the rule parameters corresponding to the generated discretized monitoring dataset are sorted using a weighted association classifier to obtain sorted rule parameters, including: The discretized monitoring dataset and state category data are combined to generate multiple rule parameters; The weighted support and confidence of the rule parameters are calculated using a weighted association classifier, and a harmonic mean is generated based on the weighted support and confidence. The rule parameters are sorted according to the harmonic mean to obtain the sorted rule parameters.

[0009] Preferably, the step of optimizing the weighted association classifier using a multi-strategy gray wolf population model to obtain the optimized weighted association classifier includes: The initial positions of the gray wolf population are generated using Henon chaotic mapping; Calculate the fitness value of each individual in the population, and select the three individuals with the best fitness as α wolf, β wolf and δ wolf respectively; The fitness function is set as the integral index of the absolute value of time multiplication error, and the convergence factor is updated using three different nonlinear curves. Calculate the control parameters based on the convergence factor, iterate the above process until the maximum number of iterations is reached, output the individual position that minimizes the fitness function value, determine the individual position with the minimum fitness function value as the optimal parameter combination of the weighted association classifier, and obtain the optimized weighted association classifier.

[0010] Preferably, the step of converting the new monitoring data into a discrete monitoring dataset and classifying it using the optimized weighted association classifier to obtain the risk level corresponding to the air spring includes: The new monitoring data is preprocessed by normalization to obtain a monitoring data vector set; The monitoring data vector set is discretized to determine the optimal number of Gaussian components, generating multiple discrete intervals to obtain the discretized test sample dataset. The discretized test sample dataset is classified using an optimized weighted association classifier. The harmonic mean of each category is calculated, and the category with the largest harmonic mean is determined as the risk level, thus obtaining the risk level corresponding to the air spring.

[0011] This invention discloses an air spring safety risk detection system, comprising: The acquisition module is used to acquire monitoring data from various sensors in the air spring and convert the monitoring data into a monitoring data vector set. The discretization module is used to divide the monitoring data vector set into multiple subsets, discretize it by fitting a Gaussian mixture model to each subset, determine the optimal number of Gaussian components, generate multiple discrete intervals, assign an interval identifier to each discrete interval, and obtain a discretized monitoring dataset. The rule sorting module is used to generate rule parameters corresponding to the discretized monitoring dataset. The rule parameters are sorted by a weighted association classifier to obtain the sorted rule parameters. The optimization module is used to optimize the weighted association classifier using a multi-strategy gray wolf population model to obtain the optimized weighted association classifier. The classification module is used to convert new monitoring data into a discrete monitoring dataset, and then classify it using the optimized weighted association classifier to obtain the risk level corresponding to the air spring.

[0012] Preferably, the acquisition module includes: The data acquisition submodule is used to acquire four-dimensional monitoring data corresponding to the air pressure sensor, displacement sensor, temperature sensor and acceleration sensor in the air spring; The normalization submodule is used to perform normalization preprocessing on the four-dimensional monitoring data to obtain a monitoring data vector set.

[0013] This invention also discloses a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described steps for detecting safety risks of air springs.

[0014] This invention also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described steps for detecting safety risks of air springs.

[0015] The embodiments of the present invention have the following advantages: In this embodiment of the invention, the air spring safety risk detection method includes: acquiring monitoring data from multiple sensors in the air spring, and converting the monitoring data into a monitoring data vector set; dividing the monitoring data vector set into multiple subsets, discretizing each subset by fitting a Gaussian mixture model, determining the optimal number of Gaussian components, generating multiple discrete intervals, assigning an interval identifier to each discrete interval, and obtaining a discretized monitoring dataset; generating rule parameters corresponding to the discretized monitoring dataset, sorting the rule parameters using a weighted association classifier, and obtaining the sorted rule parameters; optimizing the weighted association classifier using a multi-strategy gray wolf population model, and obtaining an optimized weighted association classifier; converting new monitoring data into a discretized monitoring dataset, classifying it using the optimized weighted association classifier, and obtaining the risk level corresponding to the air spring. This invention, by deploying four types of sensors—air pressure, displacement, temperature, and acceleration—achieves comprehensive perception of the multidimensional health status of the air spring, overcoming the shortcomings of existing technologies that only monitor single or a few parameters. By employing a discretization algorithm based on a Gaussian mixture model and a feature-weighted association classification algorithm, interpretable diagnostic rule parameters for "monitoring mode to risk level" are automatically extracted from historical monitoring data. This enables early prediction of potential risks in air springs, achieving a shift from "passive alarm" to "active prediction." Through multiple mechanisms, including feature weight allocation, harmonic mean sorting, and multi-strategy improvement of the Grey Wolf algorithm parameter optimization, the classification accuracy of the association classifier is significantly improved. Continuous optimization of the classifier parameters via multi-strategy improvement of the Grey Wolf algorithm allows the system to adapt to different operating conditions and environmental changes, exhibiting strong generalization ability and robustness. Attached Figure Description

[0016] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the steps of an embodiment of an air spring safety risk detection method according to the present invention. Figure 2 This is a structural block diagram of an embodiment of an air spring safety risk detection system according to an embodiment of the present invention; Figure 3 This is an internal structural diagram of a computer device according to one embodiment. Detailed Implementation

[0017] To make the technical problems, technical solutions, and beneficial effects solved by the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0018] Reference Figure 1 The diagram illustrates a flowchart of an embodiment of an air spring safety risk detection method according to the present invention, which may specifically include the following steps: Step 101: Obtain monitoring data from multiple sensors in the air spring, and convert the monitoring data into a monitoring data vector set; In this embodiment, acquiring monitoring data from multiple sensors in the air spring and converting the monitoring data into a monitoring data vector set includes: Collect four-dimensional monitoring data corresponding to the air pressure sensor, displacement sensor, temperature sensor, and acceleration sensor in the air spring; The four-dimensional monitoring data is preprocessed by normalization to obtain a monitoring data vector set.

[0019] In this embodiment of the invention, monitoring data is collected synchronously by multiple sensors deployed on the air spring. The multiple sensors include: a pressure sensor deployed inside the air spring airbag, a displacement sensor deployed on the air spring body, a temperature sensor deployed near the air spring, and an acceleration sensor deployed on the vehicle body.

[0020] Let the four-dimensional monitoring data vector collected at the i-th sampling time be: ; Where P_i represents the air pressure value at the i-th sampling time, in MPa; H_i represents the air spring height value at the i-th sampling time, in mm; T_i represents the temperature value at the i-th sampling time, in °C; and A_i represents the vibration acceleration value at the i-th sampling time, in m / s². 2 N represents the total number of samples; the superscript T indicates matrix transpose.

[0021] After data collection, the monitoring data for each dimension are preprocessed using normalization. The normalization formula is as follows: ; Where x_{i,j} represents the original value of the j-th dimension monitoring data at the i-th sampling time, μ_j is the mean of the j-th dimension monitoring data at all sampling times, and σ_j is the standard deviation of the j-th dimension monitoring data at all sampling times. These are the normalized data values. After normalization, the monitoring data vector set X = {x1, x2, …, x_N} is obtained, where x_i = [ , , , ]^T.

[0022] By simultaneously acquiring data from four types of sensors, comprehensive monitoring of four dimensions of air springs—air pressure, height, temperature, and vibration acceleration—was achieved, overcoming the limitations of existing technologies that only monitor a single or a small number of parameters. Through normalization preprocessing, the influence of different dimensions on the monitored parameters was eliminated, providing a standardized data foundation for subsequent multi-dimensional data fusion analysis, enabling parameters with different dimensions to be comprehensively processed within the same framework.

[0023] Step 102: Divide the monitoring data vector set into multiple subsets, discretize it by fitting a Gaussian mixture model to each subset, determine the optimal number of Gaussian components, generate multiple discrete intervals, assign an interval identifier to each discrete interval, and obtain the discretized monitoring dataset. Specifically, in this embodiment of the invention, the monitoring data vector set is divided into multiple subsets, and each subset is discretized by fitting a Gaussian mixture model to determine the optimal number of Gaussian components, generating multiple discrete intervals, and assigning an interval identifier to each discrete interval to obtain a discretized monitoring dataset, including: Obtain the state category data corresponding to the subset; For each subset, the expectation-maximization algorithm is used to fit the Gaussian mixture model to obtain the probability density function of the Gaussian mixture model. The probability density function of the Gaussian mixture model is then iteratively updated to determine the optimal number of Gaussian components. Calculate the mean and standard deviation of each Gaussian component to generate initial intervals. Use the analogy ratio to merge or delete overlapping initial intervals to generate multiple discrete intervals. Assign an interval identifier to each discrete interval to obtain a discretized monitoring dataset.

[0024] In this step, the continuous data in each dimension of the normalized monitoring data vector set X obtained above are discretized into multiple discrete intervals, transforming the continuous data into a discrete format suitable for association classification algorithms. Existing discretization methods (such as equal-width discretization and equal-frequency discretization) cannot preserve the multimodal distribution characteristics of the original data and will lose important data distribution information. This invention uses a discretization algorithm based on the Gaussian Mixture Model (GMM) to solve this problem.

[0025] Step 102 specifically includes the following sub-steps: Step S21: Based on the health status category C of the air spring, divide the monitoring data vector set X into |C| subsets. The health status category C (i.e., the status category data) includes: normal state, slight degradation state, moderate degradation state, severe degradation state, and fault state, for a total of 5 categories, i.e., |C| = 5. Each subset corresponds to one health status category.

[0026] Step S22: For each subset, fit a Gaussian mixture model using the Expectation Maximization (EM) algorithm. The probability density function of the Gaussian mixture model is: ; Where Θ = {w_k, μ_k, Σ_k}_{k=1}^{M} is the parameter set of the Gaussian mixture model; w_k is the weight of the k-th Gaussian component, satisfying... The value range of w_k is (0,1]; μ_k is the mean vector of the k-th Gaussian component, μ_k ∈ ^D, where D is the dimension of the monitoring data (D=4 in this invention); Σ_k is the covariance matrix of the k-th Gaussian component, Σ_k ∈ ^{D×D} is a symmetric positive definite matrix; M is the number of Gaussian components; p(x|μ_k, Σ_k) is the probability density function of the k-th Gaussian component, and its expression is: ; Where |Σ_k| represents the determinant of the covariance matrix Σ_k. Let Σ_k be the inverse of the covariance matrix Σ_k, and exp denote the natural exponential function.

[0027] The iterative update formula for the EM algorithm is as follows: ; ; ; Where t is the iteration number, and h_k^{(t)}(n) is the posterior probability that the nth sample belongs to the kth Gaussian component in the tth iteration, and its calculation formula is: ; Step S23: Determine the optimal number of Gaussian components M_opt using the Bayesian Information Criterion (BIC): ; Where L(Θ_M) is the likelihood function, representing the probability of observing the current dataset under parameter Θ_M; k_M is the number of model parameters; and N is the number of samples. The BIC criterion balances the goodness of fit and model complexity by introducing a penalty term k_M·ln N into the likelihood function, avoiding overfitting due to too many Gaussian components or underfitting due to too few Gaussian components.

[0028] Step S24: For each Gaussian component, define an initial interval [μ_{k,d} - 3σ_{k,d}, μ_{k,d} + 3σ_{k,d}] based on the 3σ principle, where μ_{k,d} is the mean of the k-th Gaussian component in the d-th dimension, and σ_{k,d} is the standard deviation of the k-th Gaussian component in the d-th dimension. The 3σ principle is based on the property of the normal distribution: approximately 99.73% of the values ​​in a normal distribution lie within the range of the mean ± 3 times the standard deviation, therefore this interval can cover the vast majority of data points.

[0029] Step S25: Merge or delete overlapping initial intervals using the Class Ratio (CR). The Class Ratio is defined as CR_c = N_c / N, where N_c is the total number of observations belonging to class c, and N is the total number of observations in the dataset. The specific rules for interval merging and deletion are as follows: (1) If two intervals belong to the same category, the two intervals are merged. The upper limit of the merged interval is the larger of the two upper limits, and the lower limit is the smaller of the two lower limits. (2) If the two intervals belong to different categories, calculate the product of the weight of the Gaussian component corresponding to each interval and the class ratio (weight × CR), retain the interval with the larger product, and delete the interval with the smaller product.

[0030] After removing all overlaps between intervals, assign a unique positive integer as the interval identifier (Interval ID) to each discrete interval.

[0031] After the above steps, each normalized monitoring data point is mapped to a discrete interval and obtains a corresponding interval identifier, thus obtaining the discrete monitoring dataset D.

[0032] The Gaussian mixture model is used to accurately model the multimodal distribution of the original monitoring data, and the BIC criterion is used to automatically determine the optimal number of Gaussian components. This ensures that the discretized data retains the multimodal distribution characteristics of the original data to the greatest extent, avoiding the defects of traditional equal-width and equal-frequency discretization methods that destroy the data distribution pattern. Supervised interval merging and deletion are performed by class ratio, which makes full use of the category label information in the discretization process, improves the adaptability of the discretization results to subsequent classification tasks, and provides high-quality discretized input data for the association classifier.

[0033] Step 103: Generate the rule parameters corresponding to the discretized monitoring dataset, and sort the rule parameters using a weighted association classifier to obtain the sorted rule parameters; In this embodiment of the invention, the rule parameters corresponding to the generation of the discretized monitoring dataset are sorted using a weighted association classifier to obtain the sorted rule parameters, including: The discretized monitoring dataset and state category data are combined to generate multiple rule parameters; The weighted support and confidence of the rule parameters are calculated using a weighted association classifier, and a harmonic mean is generated based on the weighted support and confidence. The rule parameters are sorted according to the harmonic mean to obtain the sorted rule parameters.

[0034] This step employs a weighted association classifier to generate classification association rule parameters based on the aforementioned discretized monitoring dataset D. Most existing association classification algorithms use support and confidence to generate association rule parameters, neglecting the differences in importance of attributes within specific application domains. Furthermore, using estimated minimum support and minimum confidence thresholds can prematurely delete valuable rule parameters. This invention addresses these issues by introducing feature weights and harmonic mean.

[0035] Step 103 specifically includes the following sub-steps: Step S31: Based on the degree of influence of each monitoring parameter on the safety risk of the air spring, domain experts assign characteristic weights w_P, w_H, w_T, and w_A to the air pressure value P, altitude value H, temperature value T, and vibration acceleration value A. Each weight ranges from 1 to 10, with higher values ​​indicating a greater degree of influence on safety risk. Specifically: the weight w_P for air pressure value P is 8-10 (high influence), the weight w_H for altitude value H is 7-9 (high influence), the weight w_T for temperature value T is 4-6 (medium influence), and the weight w_A for vibration acceleration value A is 4-6 (medium influence).

[0036] Step S32: For the candidate rule parameter r: Itemset → C_k, where Itemset represents one or more discrete value items of monitoring parameters included in the antecedent of the rule parameter (such as "the air pressure value is in the interval 3"), and C_k represents the category of the consequent of the rule parameter (i.e., the state category data), calculate its support: ; Where, suppcount(r) is the number of times the rule parameter r appears in dataset D, i.e., the number of samples that simultaneously satisfy both the antecedent and consequent categories of the rule; N is the total number of samples in dataset D. Support reflects the universality of the rule.

[0037] Step S33: Calculate the feature weights of the rule parameter r: ; Where w_j is the feature weight of the j-th monitoring parameter, and |Itemset(r)| is the number of monitoring parameter items included in the antecedent of the rule parameter r. The feature weight reflects the overall importance of the monitoring parameters involved in the rule parameter.

[0038] Step S34: Calculate the weighted support of the rule parameter r: ; Only rules with WeightedSupport(r) ≥ min_supp are retained, where min_supp is a preset minimum support threshold, ranging from 0.1 to 0.3. The weighted support comprehensively reflects the universality of the rule parameters and the importance of the parameters involved.

[0039] Step S35: Calculate the confidence score of the rule parameter r: ; Where atoccr(r) is the total number of times the antecedent of the rule parameter r appears in the dataset D (regardless of the consequent's class). Confidence reflects the reliability of the rule parameter, that is, the proportion of samples that satisfy the antecedent condition where the consequent class is true. Only rule parameters with Confidence(r) ≥ min_conf are retained, where min_conf is a preset minimum confidence threshold with a value of 0.5.

[0040] Step S36: Calculate the harmonic mean (HM) for each rule parameter: ; The harmonic mean is the harmonic average of weighted support and confidence, which comprehensively reflects the universality (weighted support) and reliability (confidence) of the rule parameters. It is insensitive to extreme values, avoiding the problem of distorted rule parameter evaluation due to abnormally high or low values ​​of a single indicator. The rule parameters are sorted from high to low based on the HM(r) value.

[0041] Step S37: Using the database coverage method, the sorted rule parameters are divided into a strong rule parameter set R_strong and a spare rule parameter set R_spare. The specific process of the database coverage method is as follows: Each rule parameter is examined sequentially according to its HM value from high to low. If the rule parameter can correctly cover samples in the training dataset that have not yet been covered by existing rules, then the rule parameter is added to the strong rule parameter set R_strong, and the samples covered by the rule parameter are marked. If the rule parameter cannot correctly classify any uncovered samples, then it is added to the spare rule parameter set R_spare. The above process is repeated until all training samples are covered or all rule parameters have been checked.

[0042] By introducing a feature weighting mechanism, monitoring parameters with a greater impact on safety risks (such as air pressure and altitude) are given higher priority in the rule parameter generation process, improving the quality and application relevance of the rule parameters. By introducing the harmonic mean (HM) as the ranking index for rule parameters, the weighted support and confidence of the rule parameters are comprehensively balanced, avoiding the shortcomings of traditional methods that prematurely delete valuable rule parameters by using estimated minimum support and minimum confidence thresholds. By using a database coverage method to stratify the rule parameters into strong rule sets and backup rule sets, a dual guarantee mechanism is provided for subsequent classification decisions, improving the robustness of classification.

[0043] Step 104: Optimize the weighted association classifier using a multi-strategy gray wolf population model to obtain the optimized weighted association classifier; In practical applications, the optimization of the weighted association classifier using a multi-strategy gray wolf population model to obtain the optimized weighted association classifier includes: The initial positions of the gray wolf population are generated using Henon chaotic mapping; Calculate the fitness value of each individual in the population, and select the three individuals with the best fitness as α wolf, β wolf and δ wolf respectively; The fitness function is set as the integral index of the absolute value of time multiplication error, and the convergence factor is updated using three different nonlinear curves. Calculate the control parameters based on the convergence factor, iterate the above process until the maximum number of iterations is reached, output the individual position that minimizes the fitness function value, determine the individual position with the minimum fitness function value as the optimal parameter combination of the weighted association classifier, and obtain the optimized weighted association classifier.

[0044] This step employs a multi-strategy improvement to the Grey Wolf algorithm, globally optimizing the parameters of the aforementioned association classification algorithm (including the minimum support threshold min_supp, the minimum confidence threshold min_conf, and the weight coefficients in rule ranking). The conventional Grey Wolf algorithm (GWO) suffers from drawbacks such as imbalance between global and local optimization, slow convergence speed, and reliance on the initial solution. This invention addresses these problems through three improvement strategies.

[0045] Step 104 specifically includes the following sub-steps: Step S41: The initial positions of the gray wolf population are generated using Henon chaotic mapping to enhance population diversity and randomness. The Henon chaotic mapping equation is: ; ; Where x_n and y_n are state variables during the iteration process; a and b are control parameters. When a = 1.4 and b = 0.3, the system is in a hyperchaotic state, and the generated chaotic sequence has good randomness and ergodicity. The generated chaotic sequence is used to initialize the position vector of each individual in the gray wolf population. The position vector of each individual represents a set of parameters for the association classification algorithm to be optimized.

[0046] Step S42: Calculate the fitness value of each individual in the population, and select the three individuals with the best fitness as α wolf, β wolf, and δ wolf, respectively. The fitness function adopts the integral of time multiplied by the absolute value of error (ITAE) index defined in step S46.

[0047] Step S43: Update the convergence factors a1, a2, and a3 using three different nonlinear curves, instead of the linear decreasing method of the convergence factor in the traditional GWO algorithm. ; ; ; Where a_max = 2 is the maximum value of the convergence factor, a_min = 0 is the minimum value of the convergence factor, t is the current iteration number, and T is the maximum iteration number. The three different nonlinear curves correspond to the convergence factors of the α wolf, β wolf, and δ wolf, respectively, so that the three alpha wolves have different convergence characteristics during the search process, thereby better balancing global search and local optimization.

[0048] Step S44: Calculate the control parameters based on the convergence factor: ; Where r_i is a random number uniformly distributed between [0,1], and i = 1, 2, 3 correspond to the control parameters of α wolf, β wolf, and δ wolf, respectively. When |A_i| ≥ 1, the algorithm tends to perform a global search; when |A_i| < 1, the algorithm tends to perform a local optimization.

[0049] Step S45: Introduce the position information sharing mechanism of Particle Swarm Optimization (PSO) and update the velocity and position of the individual gray wolves according to the following formula: ; ; Where V(t) is the velocity vector of the individual gray wolf at the t-th iteration; ω is the inertia weight, with a value range of [0.4, 0.9]; C1, C2, and C3 are the learning factors for α, β, and δ wolves, respectively, with a value range of [1.2, 3.0]; r1, r2, and r3 are random numbers uniformly distributed between [0, 1]; X_α, X_β, and X_δ are the position vectors of α, β, and δ wolves, respectively; and X(t) is the position vector of the individual gray wolf at the t-th iteration. This velocity update formula integrates the mechanism of particle movement towards the individual optimal and global optimal directions in the PSO algorithm, enabling the individual gray wolf to not only follow the guidance of the three alpha wolves but also comprehensively consider its own historical movement trend, thus accelerating the convergence speed of the algorithm.

[0050] Step S46: Use the integral of time-weighted absolute error (ITAE) as the fitness function: ; Where t is the system runtime and e(t) is the classification error of the classifier. The ITAE metric, by multiplying the error by a time weight, penalizes later errors more severely, thus providing a comprehensive evaluation of the classifier's accuracy and response speed.

[0051] Iteratively execute steps S42 to S45 until the maximum number of iterations T is reached, and output the individual position that minimizes the fitness function value, i.e., the optimal combination of association classifier parameters.

[0052] A specific example is shown below. For ease of understanding, the initial algorithm parameters can be set as follows: population size N_pop = 30, maximum number of iterations T = 100, maximum inertia weight ω_max = 0.9, minimum inertia weight ω_min = 0.4, individual learning factors C1_max = 3.0, C1_min = 1.2, group learning factors C2_max = 3.0, C2_min = 1.2, maximum convergence factor a_max = 2, minimum convergence factor a_min = 0.

[0053] The Henon chaotic map (a = 1.4, b = 0.3) is used to generate the position vectors of 30 initial individuals. The position vector of each individual represents a set of parameters to be optimized (including min_supp, min_conf and weight coefficients in the rule sorting).

[0054] Calculate the fitness value (ITAE value) of each individual, and select the three individuals with the best fitness as α wolf, β wolf and δ wolf.

[0055] The convergence factor is updated according to three nonlinear curves: a1 = 2 - 2 × (t / 100) 3 ; a2= (2 - 2 × t / 100) × (1 - (t / 100) 3 ) ; a3 = 2 × (1 - (t / 100)) 2 ) ; Calculate the control parameters A_i = 2a_i × r_i - a_i (i = 1, 2, 3).

[0056] Update the speed and location of each individual gray wolf according to the PSO information sharing mechanism: V(t+1) = ω·V(t) + C1·r1·(X_α - X(t)) + C2·r2·(X_β - X(t)) + C3·r3·(X_δ - X(t)); X(t+1) = X(t) + V(t+1); The iterations are repeated until the maximum number of iterations (100) is reached. The output is the location of the individual that minimizes the ITAE value, i.e., the optimal parameter combination. In this embodiment, the optimized parameter combination of the association classifier is: min_supp = 0.12, min_conf = 0.48.

[0057] Initializing the population using Henon chaotic mapping improves the quality of initial solutions and population diversity, preventing the algorithm from getting trapped in local optima. Updating the convergence factor using three different nonlinear curves allows different alpha wolves to exhibit differentiated convergence characteristics during the search process, better balancing global search and local optimization. Introducing a PSO position information sharing mechanism accelerates the algorithm's convergence speed. The ITAE fitness function achieves comprehensive optimization of classifier accuracy and response speed. Compared to the traditional GWO algorithm, the PSO_IGWO algorithm of this invention significantly improves both optimization accuracy and convergence speed, and is less prone to getting trapped in local optima.

[0058] Step 105: Convert the new monitoring data into a discrete monitoring dataset, and classify it using the optimized weighted association classifier to obtain the risk level corresponding to the air spring.

[0059] Specifically, the process of converting the new monitoring data into a discrete monitoring dataset and classifying it using the optimized weighted association classifier to obtain the risk level corresponding to the air spring includes: The new monitoring data is preprocessed by normalization to obtain a monitoring data vector set; The monitoring data vector set is discretized to determine the optimal number of Gaussian components, generating multiple discrete intervals to obtain the discretized test sample dataset. The discretized test sample dataset is classified using an optimized weighted association classifier. The harmonic mean of each category is calculated, and the category with the largest harmonic mean is determined as the risk level, thus obtaining the risk level corresponding to the air spring.

[0060] This step involves normalizing the monitoring data of the air springs to be tested, then discretizing it, and finally determining the risk level based on the generated set of rule parameters and outputting a warning signal.

[0061] Step 105 specifically includes the following sub-steps: Step S51: Preprocess the monitoring data to be detected according to the normalization formula in step S1 to obtain the normalized monitoring data vector x_test = [ , , , ]^T.

[0062] Step S52: Based on the discrete interval division result determined in step S2, map each dimension of data to the corresponding discrete interval to obtain the corresponding interval identifier, and obtain the discretized test sample.

[0063] Step S53: In the strong rule parameter set R_strong, find all rule parameters whose antecedents match the test sample (i.e., all items in the antecedents of the rule parameters are consistent with the discrete values ​​of the test sample). If matching rule parameters exist, group the matching rule parameters by category (health status category), calculate the average HM value for each category, and select the category with the largest average HM value as the predicted risk level.

[0064] Step S54: If there is no matching rule parameter in the strong rule parameter set R_strong, then search for all rule parameters in the spare rule parameter set R_spare that match the antecedent with the test sample, and perform voting judgment in the same way as in step S53.

[0065] Step S55: If there are no matching rule parameters in either the strong rule parameter set R_strong or the spare rule parameter set R_spare, the test sample is assigned to the default category, which is the risk level that appears most frequently in the training data (usually "normal state").

[0066] Step S56: Output the corresponding warning signal based on the judgment result. The warning signal includes four levels, as shown in the table below:

[0067] Table 1: Risk Levels This invention achieves comprehensive perception of the multidimensional health status of air springs by deploying four types of sensors: air pressure, displacement, temperature, and acceleration, overcoming the shortcomings of existing technologies that only monitor single or a few parameters. Through a discretization algorithm based on a Gaussian mixture model and a feature-weighted association classification algorithm, it automatically extracts interpretable diagnostic rule parameters from historical monitoring data, from "monitoring mode to risk level," enabling early prediction of potential risks to air springs and achieving a shift from "passive alarm" to "active prediction." Through multiple mechanisms, including feature weight allocation, harmonic mean sorting, and multi-strategy improvement of the Grey Wolf algorithm parameter optimization, the classification accuracy of the association classifier is significantly improved. Continuous optimization of the classifier parameters through multi-strategy improvement of the Grey Wolf algorithm enables the system to adapt to different operating conditions and environmental changes, exhibiting strong generalization ability and robustness.

[0068] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0069] Reference Figure 2 The diagram illustrates a structural block diagram of an air spring safety risk detection system according to an embodiment of the present invention, which may specifically include the following modules: The acquisition module 301 is used to acquire monitoring data from multiple sensors in the air spring and convert the monitoring data into a monitoring data vector set. Discretization module 302 is used to divide the monitoring data vector set into multiple subsets, discretize it by fitting a Gaussian mixture model to each subset, determine the optimal number of Gaussian components, generate multiple discrete intervals, assign an interval identifier to each discrete interval, and obtain a discretized monitoring dataset. The rule sorting module 303 is used to generate rule parameters corresponding to the discretized monitoring dataset, and sort the rule parameters by a weighted association classifier to obtain the sorted rule parameters. Optimization module 304 is used to optimize the weighted association classifier using a multi-strategy gray wolf population model to obtain an optimized weighted association classifier; The classification module 305 is used to convert new monitoring data into a discrete monitoring dataset, and classify it through the optimized weighted association classifier to obtain the risk level corresponding to the air spring.

[0070] Preferably, the acquisition module includes: The data acquisition submodule is used to acquire four-dimensional monitoring data corresponding to the air pressure sensor, displacement sensor, temperature sensor and acceleration sensor in the air spring; The normalization submodule is used to perform normalization preprocessing on the four-dimensional monitoring data to obtain a monitoring data vector set.

[0071] Preferably, the discretization module includes: The status acquisition submodule is used to acquire the status category data corresponding to the subset; The fitting submodule is used to fit a Gaussian mixture model to each subset using the expectation-maximization algorithm, obtain the probability density function of the Gaussian mixture model, and iteratively update the probability density function of the Gaussian mixture model to determine the optimal number of Gaussian components. The interval generation submodule is used to calculate the mean and standard deviation of each Gaussian component, generate initial intervals, merge or delete overlapping initial intervals using analogy ratios, generate multiple discrete intervals, assign interval identifiers to each discrete interval, and obtain a discretized monitoring dataset.

[0072] Preferably, the rule sorting module includes: The rule generation submodule is used to combine the discretized monitoring dataset and state category data to generate multiple rule parameters; The weighted calculation submodule is used to calculate the weighted support and confidence of the rule parameters through a weighted association classifier, and generate a harmonic mean based on the weighted support and confidence. The sorting submodule is used to sort the rule parameters according to the harmonic mean, so as to obtain the sorted rule parameters.

[0073] Preferably, the optimization module includes: The initialization submodule is used to generate the initial positions of the gray wolf population using Henon chaotic mapping; The fitness submodule is used to calculate the fitness value of each individual in the population, and select the three individuals with the best fitness as α wolf, β wolf and δ wolf respectively; The convergence factor submodule is used to set the time multiplication error absolute value integral index as the fitness function and update the convergence factor using three different nonlinear curves. The iterative submodule is used to calculate the control parameters based on the convergence factor, iterate until the maximum number of iterations is reached, and output the individual position that minimizes the fitness function value. The individual position with the minimum fitness function value is determined as the optimal combination of parameters for the weighted association classifier, thus obtaining the optimized weighted association classifier.

[0074] Preferably, the classification module includes: The preprocessing submodule is used to normalize and preprocess the new monitoring data to obtain a monitoring data vector set. The discretization submodule is used to discretize the monitoring data vector set, determine the optimal number of Gaussian components, generate multiple discrete intervals, and obtain the discretized test sample dataset. The prediction submodule is used to classify the discretized test sample dataset using an optimized weighted association classifier, calculate the harmonic mean of each class, determine the class with the largest harmonic mean as the risk level, and obtain the risk level corresponding to the air spring.

[0075] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0076] Specific limitations regarding the air spring safety risk detection system can be found in the limitations of the air spring safety risk detection method described above, and will not be repeated here. Each module in the aforementioned air spring safety risk detection system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0077] The air spring safety risk detection system provided above can be used to execute the air spring safety risk detection method provided in any of the above embodiments, and has the corresponding functions and beneficial effects.

[0078] In one embodiment, a computer device is provided, the internal structure of which can be shown as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices 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 the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for detecting safety risks of an air spring. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0079] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0080] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the steps described above.

[0081] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, performs the steps described above.

[0082] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0083] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0084] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of apparatus, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction methods implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0088] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0089] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or terminal device that includes said element.

[0090] The present invention has provided a detailed description of an air spring safety risk detection method, an air spring safety risk detection system, a computer device, and a storage medium. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for detecting safety risks of air springs, characterized in that, include: Acquire monitoring data from multiple sensors in the air spring, and convert the monitoring data into a monitoring data vector set; The monitoring data vector set is divided into multiple subsets. A Gaussian mixture model is fitted to each subset for discretization. The optimal number of Gaussian components is determined, multiple discrete intervals are generated, and an interval identifier is assigned to each discrete interval to obtain the discretized monitoring dataset. Generate rule parameters corresponding to the discrete monitoring dataset, and sort the rule parameters using a weighted association classifier to obtain the sorted rule parameters; The weighted association classifier is optimized by using a multi-strategy gray wolf population model to obtain an optimized weighted association classifier. The new monitoring data is converted into a discrete monitoring dataset, which is then classified using the optimized weighted association classifier to obtain the risk level corresponding to the air spring.

2. The method according to claim 1, characterized in that, The process of acquiring monitoring data from multiple sensors in the air spring and converting the monitoring data into a monitoring data vector set includes: Collect four-dimensional monitoring data corresponding to the air pressure sensor, displacement sensor, temperature sensor, and acceleration sensor in the air spring; The four-dimensional monitoring data is preprocessed by normalization to obtain a monitoring data vector set.

3. The method according to claim 1, characterized in that, The monitoring data vector set is divided into multiple subsets, and a Gaussian mixture model is fitted to each subset for discretization. The optimal number of Gaussian components is determined, multiple discrete intervals are generated, and an interval identifier is assigned to each discrete interval to obtain a discretized monitoring dataset, including: Obtain the state category data corresponding to the subset; For each subset, the expectation-maximization algorithm is used to fit the Gaussian mixture model to obtain the probability density function of the Gaussian mixture model. The probability density function of the Gaussian mixture model is then iteratively updated to determine the optimal number of Gaussian components. Calculate the mean and standard deviation of each Gaussian component to generate initial intervals. Use the analogy ratio to merge or delete overlapping initial intervals to generate multiple discrete intervals. Assign an interval identifier to each discrete interval to obtain a discretized monitoring dataset.

4. The method according to claim 1, characterized in that, The rule parameters corresponding to the generated discretized monitoring dataset are sorted using a weighted association classifier to obtain the sorted rule parameters, including: The discretized monitoring dataset and state category data are combined to generate multiple rule parameters; The weighted support and confidence of the rule parameters are calculated using a weighted association classifier, and a harmonic mean is generated based on the weighted support and confidence. The rule parameters are sorted according to the harmonic mean to obtain the sorted rule parameters.

5. The method according to claim 1, characterized in that, The optimization of the weighted association classifier using a multi-strategy gray wolf population model to obtain the optimized weighted association classifier includes: The initial positions of the gray wolf population are generated using Henon chaotic mapping; Calculate the fitness value of each individual in the population, and select the three individuals with the best fitness as α wolf, β wolf and δ wolf respectively; The fitness function is set as the integral index of the absolute value of time multiplication error, and the convergence factor is updated using three different nonlinear curves. Calculate the control parameters based on the convergence factor, iterate the above process until the maximum number of iterations is reached, output the individual position that minimizes the fitness function value, determine the individual position with the minimum fitness function value as the optimal parameter combination of the weighted association classifier, and obtain the optimized weighted association classifier.

6. The method according to claim 1, characterized in that, The process of converting the new monitoring data into a discrete monitoring dataset and classifying it using the optimized weighted association classifier to obtain the risk level corresponding to the air spring includes: The new monitoring data is preprocessed by normalization to obtain a monitoring data vector set; The monitoring data vector set is discretized to determine the optimal number of Gaussian components, generating multiple discrete intervals to obtain the discretized test sample dataset. The discretized test sample dataset is classified using an optimized weighted association classifier. The harmonic mean of each category is calculated, and the category with the largest harmonic mean is determined as the risk level, thus obtaining the risk level corresponding to the air spring.

7. An air spring safety risk detection system, characterized in that, include: The acquisition module is used to acquire monitoring data from various sensors in the air spring and convert the monitoring data into a monitoring data vector set. The discretization module is used to divide the monitoring data vector set into multiple subsets, discretize it by fitting a Gaussian mixture model to each subset, determine the optimal number of Gaussian components, generate multiple discrete intervals, assign an interval identifier to each discrete interval, and obtain a discretized monitoring dataset. The rule sorting module is used to generate rule parameters corresponding to the discretized monitoring dataset. The rule parameters are sorted by a weighted association classifier to obtain the sorted rule parameters. The optimization module is used to optimize the weighted association classifier using a multi-strategy gray wolf population model to obtain the optimized weighted association classifier. The classification module is used to convert new monitoring data into a discrete monitoring dataset, and then classify it using the optimized weighted association classifier to obtain the risk level corresponding to the air spring.

8. The system according to claim 1, characterized in that, The acquisition module includes: The data acquisition submodule is used to acquire four-dimensional monitoring data corresponding to the air pressure sensor, displacement sensor, temperature sensor and acceleration sensor in the air spring; The normalization submodule is used to perform normalization preprocessing on the four-dimensional monitoring data to obtain a monitoring data vector set.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the air spring safety risk detection method according to 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 the processor, it implements the steps of the air spring safety risk detection method according to any one of claims 1 to 7.

Images