An intelligent axle monitoring system

Abstract

The present invention provides an intelligent axle monitoring system, which relates to the technical field of vehicle monitoring, and includes a monitoring point selection module, a data acquisition module, a current damage determination module, a remaining life calculation module, a model training module, and an alarm module; the monitoring point selection module screens the monitoring point set to obtain multiple monitoring points; the data acquisition module acquires the current stress sequences at multiple monitoring points; the current damage determination module is used to process the current stress sequences to obtain the current damage of each monitoring point; the remaining life calculation module processes the current damage and historical damage through an RBF model to obtain the remaining life of the axle; the model training module is used to train a genetic algorithm and an RBF model, and the genetic algorithm and the RBF model are obtained through joint training; the alarm module is used to issue an alarm when the remaining life does not meet the safety threshold; so as to improve the accuracy of axle life prediction, more accurately obtain the life condition of the axle, and avoid traffic accidents.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle monitoring, and more particularly, to an intelligent vehicle bridge monitoring system. Background Art

[0002] The vehicle bridge is an important part of the vehicle chassis, used to connect the left and right wheels of the vehicle and support the weight of the vehicle body. It is one of the core components of the vehicle. Classified by position, the vehicle bridge can be divided into the front axle and the rear axle, etc.; classified by function, the vehicle bridge can be divided into the drive axle, the driven axle and the support axle, etc. The modern vehicle bridge monitoring technology mainly relies on a variety of sensors (such as strain gauges, acceleration sensors, temperature sensors and lubricant quality sensors, etc.) to monitor the vehicle bridge comprehensively. The system monitors key parameters including axle load, vibration frequency, temperature change and stress distribution through real-time data acquisition and wireless transmission. An intelligent analysis system is used for fault warning, life assessment and anomaly detection, and real-time monitoring and management are realized through in-vehicle display and remote monitoring platform. However, when evaluating the life of the vehicle bridge in the prior art, it often focuses on the maximum load-bearing point of the vehicle bridge, while ignoring the damage of other parts. As a whole component of the vehicle, there is also a risk of damage to other parts. Therefore, only monitoring and analyzing the load-bearing point makes the monitoring blind area of the vehicle bridge too large, and there is a large deviation between the life prediction and the actual life of the vehicle bridge.

[0003] In view of this, the present application proposes an intelligent vehicle bridge monitoring system, which monitors multiple points of the vehicle bridge and processes the monitoring data of multiple points to obtain the remaining life of the vehicle bridge, improves the accuracy of the vehicle bridge life prediction, enables a more accurate acquisition of the life condition of the vehicle bridge, and makes a feedback in time according to the predicted life to avoid traffic accidents. Summary of the Invention

[0004] The object of the present invention is to provide an intelligent monitoring system for an axle, including a monitoring point selection module, a data acquisition module, a current damage determination module, a remaining life calculation module, a model training module, and an alarm module; the monitoring point selection module screens a set of monitoring points through a genetic algorithm to obtain a plurality of monitoring positions; the data acquisition module includes a plurality of stress sensors, and by setting the stress sensors at the plurality of monitoring positions, the current stress sequences at the plurality of monitoring positions are obtained; the current damage determination module is used to process the current stress sequences to obtain the current damage of each monitoring position; the remaining life calculation module processes the current damage and historical damage of the plurality of monitoring positions through an RBF (Radial Basis Function) neural network model to obtain the remaining life of the axle; the model training module is used to train the genetic algorithm and the RBF model, and the genetic algorithm and the RBF model are obtained through joint training, and the purpose of the joint training is to maximize the objective function, and the objective function is: ; Wherein, represents the objective function; , and respectively represent the first objective parameter, the second objective parameter, and the third objective parameter; i represents the monitoring point variable; I represents the total number of monitoring points; represents the remaining life output by the RBF model obtained from the previous round of training; t represents the time variable of the obtained monitoring point; T represents the total number of time points of the data at the obtained monitoring point; represents the actual remaining life at time point t; represents the predicted remaining life at time point t output by the RBF model obtained from the current round of training; represents the weight of the jth RBF neuron; the alarm module is used to issue an alarm when the remaining life does not meet the safety threshold.

[0005] Further, the monitoring point selection module includes a construction unit, a screening unit, and a loop unit; the construction unit constructs a plurality of monitoring position selection vectors based on the set of monitoring points; the monitoring position selection vector is used to represent the selection situation of the monitoring positions in the set of monitoring points; the screening unit calculates the fitness value of each monitoring position selection vector and determines a plurality of target monitoring position selection vectors; the fitness value is related to the sensitivity of the monitoring position to the remaining life; the loop unit is used to process the target monitoring position selection vectors to obtain a plurality of new monitoring position selection vectors, and repeat the screening operation until the fitness value converges.

[0006] Further, the fitness function for solving the fitness value is: ; where represents the fitness value; i represents the monitoring point variable; I represents the total number of monitoring points in the monitoring point set; represents the selection situation of the i-th monitoring point; represents the position of the i-th monitoring point on the axle; represents the remaining life of the axle; represents the sensitivity of the monitoring point to the remaining life.

[0007] Furthermore, the current damage determination module includes a stress sequence decomposition unit, a cycle number determination unit, an equivalent stress calculation unit, and a current damage determination unit; the stress sequence decomposition unit is used to process the current stress sequence to obtain a stress cycle set and the cycle characteristics of each stress cycle; the cycle characteristics include stress amplitude, mean stress, and total cycle number; the cycle number determination unit determines the current cycle number at each stress level based on the stress amplitude; the equivalent stress calculation unit determines the equivalent stress based on the current cycle number and the stress level; the current damage determination unit determines the current damage based on the equivalent stress and the current cycle number.

[0008] Furthermore, the current stress sequence is processed by the rain flow counting method to obtain the stress cycle set and the cycle characteristics.

[0009] Furthermore, the stress range of each stress level is: ; ; ; where represents the stress of the k-th stress level; and represent the minimum value and the maximum value of the k-th stress respectively; represents the stress value when the vehicle is unloaded; k represents the stress level variable; represents the total number of stress levels; represents the stress value when the vehicle is fully loaded; the calculation formula for the equivalent stress is: ; where represents the equivalent stress of the k-th stress; represents the actual amplitude of the k-th stress; represents the mean stress of the k-th stress; the current damage is: ; where represents the current damage; represents the actual cycle number of the k-th stress; represents the first material parameter; m represents the second material parameter.

[0010] Further, the remaining life is: ; where L represents the predicted remaining life; j represents the hidden layer node variable; J represents the number of hidden layer nodes; represents the weight of the j-th RBF neuron; exp represents the exponential function; i represents the monitoring point variable; I represents the total number of monitoring points; represents the influence factor of each monitoring point on the remaining life; represents the damage of the i-th monitoring point at the t-th time point; represents the i-th component of the j-th RBF center point; represents the expansion constant.

[0011] Further, the model training module further includes a training acceleration unit, and the training acceleration unit includes an initial model determination layer, a genetic algorithm update layer, an RBF model update layer, a first loop layer, an update acceleration layer, and a second loop layer; the initial model determination layer determines an initial RBF model based on the initial monitoring point influence factor and the initial neuron weight; the genetic algorithm update layer updates the initial genetic algorithm based on the initial RBF model to obtain a first genetic algorithm and a first monitoring point influence factor; the RBF model update layer updates the initial RBF model based on the first monitoring point influence factor to obtain a first RBF model and a first neuron weight; the first loop layer is used to repeat the update operations of the genetic algorithm and the RBF model until multiple monitoring point influence factors and multiple neuron weights are obtained; the update acceleration layer is used to process the multiple monitoring point influence factors and the multiple neuron weights to obtain accelerated monitoring point influence factors and accelerated neuron weights; the second loop layer replaces the initial monitoring point influence factor and the initial neuron weight with the accelerated monitoring point influence factor and the accelerated neuron weight respectively, and repeats the update operations of the genetic algorithm and the RBF model until the fitness value of the genetic algorithm converges.

[0012] Further, there are three of the multiple monitoring point influence factors and the multiple neuron weights, and the obtaining of the accelerated monitoring point influence factors and the accelerated neuron weights is: ; where represents the accelerated monitoring point influence factor; represents the accelerated neuron weight; , and represent the third, second, and first monitoring point influence factors respectively, and the fitness values of the third, second, and first monitoring point influence factors decrease in sequence; , and represent the third, second, and first neuron weights respectively, and the objective function values corresponding to the third, second, and first neuron weights decrease in sequence.

[0013] Further, the update acceleration layer also determines an acceleration objective function value based on the acceleration monitoring point influence factor and the acceleration neuron weight, and compares the acceleration objective function value with a third objective function value; when the acceleration objective function value is greater than or equal to the third objective function value, the second recurrent layer replaces the initial monitoring point influence factor and the initial neuron weight with the acceleration monitoring point influence factor and the acceleration neuron weight respectively; otherwise, the second recurrent layer replaces the initial monitoring point influence factor and the initial neuron weight with the third monitoring point influence factor and the third neuron weight respectively.

[0014] The technical solution of the embodiment of the present invention has at least the following advantages and beneficial effects:

[0015] In the present invention, a genetic algorithm is used to screen a set of monitoring points to obtain multiple monitoring points, so as to obtain the spatial variation of the stress of the axle, and the state of the whole axle is reflected by a certain number of monitoring points, avoiding the uncertainty brought by artificial selection.

[0016] In the present invention, an RBF model is used to process the damage at multiple time points, so as to obtain the temporal variation of the axle life, and the remaining life of the axle is determined according to the potential connection of the damage at multiple time points, improving the prediction accuracy of the axle life.

[0017] The model training module of the present invention updates the genetic algorithm and the RBF model through joint training, can connect the axle in time and space, updates the RBF model through the result of the genetic algorithm, and then updates the genetic algorithm through the updated result of the RBF model, enabling the RBF model to better process the result output by the genetic algorithm, optimizing resource allocation, reducing the noise of data transmission and processing between models, and improving the prediction accuracy.

[0018] The training acceleration unit provided by the present invention can improve the speed of model training, reduce the number of training times, and reduce energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is an exemplary schematic diagram of an axle intelligent monitoring system provided by the present invention;

[0020] Figure 2 It is an exemplary schematic diagram of the monitoring point selection module provided by the present invention;

[0021] Figure 3 It is an exemplary schematic diagram of the current damage determination module provided by the present invention;

[0022] Figure 4 It is an exemplary schematic diagram of the training acceleration unit provided by the present invention;

[0023] Icons: 1 - Monitoring point selection module, 2 - Data acquisition module, 3 - Current damage determination module, 4 - Remaining life calculation module, 5 - Model training module, 6 - Alarm module, 11 - Construction unit, 12 - Screening unit, 13 - Circulation unit, 31 - Stress sequence decomposition unit, 32 - Cycle number determination unit, 33 - Equivalent stress calculation unit, 34 - Current damage determination unit, 311 - Initial model determination layer, 312 - Genetic algorithm update layer, 313 - RBF model update layer, 314 - First circulation layer, 315 - Update acceleration layer, 316 - Second circulation layer. Specific implementation manners

[0024] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0025] Figure 1 An exemplary schematic diagram of an axle intelligent monitoring system provided by the present invention. As Figure 1 shown, the axle intelligent monitoring system provided by the present invention includes a monitoring point selection module 1, a data acquisition module 2, a current damage determination module 3, a remaining life calculation module 4, a model training module 5, and an alarm module 6.

[0026] The monitoring point selection module 1 screens the monitoring point set through a genetic algorithm to obtain multiple monitoring point positions. A monitoring point refers to a point for monitoring the axle. Monitoring points can be set at positions such as leaf spring seats, axle housings, axle tubes, bearing seats, and brake camshaft brackets. A monitoring point position refers to the position of a monitoring point. For more content about the monitoring point selection module, see Figure 2 and its related descriptions.

[0027] The data acquisition module 2 includes multiple stress sensors. By setting the stress sensors at the multiple monitoring point positions, the current stress sequence at the multiple monitoring point positions is obtained. The current stress sequence may refer to the stress values arranged in chronological order obtained in the current period. Each monitoring point position corresponds to a current stress sequence.

[0028] The current damage determination module 3 is used to process the current stress sequence to obtain the current damage of each monitoring point position. The current damage may refer to the damage value of the current axle. For more content about the current damage determination module, see Figure 3 and its related descriptions.

[0029] The remaining life calculation module 4 processes the current damage and historical damage of the multiple monitoring points through the RBF model to obtain the remaining life of the axle. The historical damage refers to the damage value of the axle obtained in the past. The remaining life refers to the time or kilometers that the axle can still be used. The remaining life is:

[0030] ;

[0031] where L represents the predicted remaining life; j represents the hidden layer node variable; J represents the number of hidden layer nodes; represents the weight of the j-th RBF neuron; exp represents the exponential function; i represents the monitoring point variable; I represents the total number of monitoring points; represents the influence factor of each monitoring point on the remaining life; represents the damage of the i-th monitoring point at the t-th time point; represents the i-th component of the j-th RBF center point; represents the expansion constant.

[0032] The model training module 5 is used to train the genetic algorithm and the RBF model. The genetic algorithm and the RBF model are obtained through joint training. The purpose of the joint training is to maximize the objective function. The objective function is:

[0033] ;

[0034] where represents the objective function; , and respectively represent the first objective parameter, the second objective parameter, and the third objective parameter; i represents the monitoring point variable; I represents the total number of monitoring points; represents the remaining life output by the RBF model obtained from the previous round of training; t represents the time variable of the obtained monitoring point; T represents the total number of time points of the data at the obtained monitoring point; represents the actual remaining life at time point t; represents the predicted remaining life at time point t output by the RBF model obtained from the current round of training; represents the weight of the j-th RBF neuron. For more content about the model training module, see Figure 4 and its related descriptions.

[0035] The alarm module is used to issue an alarm when the remaining life does not meet the safety threshold. The safety threshold can refer to the minimum remaining life value set in advance. For example, an alarm is issued when the remaining life is less than 1 week.

[0036] Figure 2An exemplary schematic diagram of the monitoring point selection module provided by the present invention. As Figure 2 shown, the monitoring point selection module 1 includes a construction unit 11, a screening unit 12, and a loop unit 13.

[0037] The construction unit constructs a plurality of monitoring point selection vectors based on the monitoring point set; the monitoring point selection vector is used to represent the selection situation of the monitoring points in the monitoring point set, with the selected value being 1 and the unselected value being 0.

[0038] The screening unit calculates the fitness value of each monitoring point selection vector and determines a plurality of target monitoring point selection vectors; the fitness value is related to the sensitivity of the monitoring point to the remaining life. Sort the monitoring point selection vectors according to the fitness value, and use a preset number or the monitoring point selection vectors with fitness values greater than the preset fitness threshold as the target monitoring point selection vectors. The fitness function for solving the fitness value is:

[0039] ;

[0040] Wherein, represents the fitness value; i represents the monitoring point variable; I represents the total number of monitoring points in the monitoring point set; represents the selection situation of the i-th monitoring point; represents the position of the i-th monitoring point on the axle; represents the remaining life of the axle; represents the sensitivity of the monitoring point to the remaining life.

[0041] The loop unit is used to process the target monitoring point selection vectors to obtain a plurality of new monitoring point selection vectors, and repeat the screening operation until the fitness value converges. The convergence of the fitness value may refer to that the fitness difference between the individual with the optimal fitness value obtained in the previous loop and the individual with the optimal fitness value obtained in the current loop is less than the preset loop difference threshold. The preset loop difference threshold is the threshold condition for terminating the loop of the genetic algorithm. The condition for terminating the loop can also be in other forms, for example, the loop reaches the maximum number of loops.

[0042] Figure 3 An exemplary schematic diagram of the current damage determination module provided by the present invention. As Figure 3 shown, the current damage determination module 3 includes a stress sequence decomposition unit 31, a cycle number determination unit 32, an equivalent stress calculation unit 33, and a current damage determination unit 34.

[0043] The stress sequence decomposition unit is used to process the current stress sequence to obtain a stress cycle set and the cycle characteristics of each stress cycle; the cycle characteristics include stress amplitude, mean stress, and total number of cycles. In some embodiments, the current stress sequence can be processed by the rain flow counting method to obtain the stress cycle set and the cycle characteristics. A stress cycle can refer to the cyclic pattern of stress variation over time when the axle is subjected to vehicle static and dynamic loads. Stress amplitude refers to the maximum variation amplitude of stress in a stress cycle. Mean stress refers to the average value of stress in a stress cycle. The total number of cycles refers to the total number of stress cycles in the stress cycle set.

[0044] The cycle number determination unit determines the current cycle number at each stress level based on the stress amplitude. The current cycle number can refer to the number of stress cycles at each stress level in the current sampling period. The stress range for each stress level is:

[0045] ;

[0046] ;

[0047] ;

[0048] where represents the stress of the k-th stress level; and represent the minimum and maximum values of the k-th stress respectively; represents the stress value when the vehicle is unloaded; k represents the stress level variable; represents the total number of stress levels; represents the stress value when the vehicle is fully loaded.

[0049] The equivalent stress calculation unit determines the equivalent stress based on the current cycle number and the stress level. The calculation formula for the equivalent stress is:

[0050] ;

[0051] where represents the equivalent stress of the k-th stress; represents the actual amplitude of the k-th stress; represents the mean stress of the k-th stress.

[0052] The current damage determination unit determines the current damage based on the equivalent stress and the current cycle number. The current damage is:

[0053] ;

[0054] where Indicates the current damage; Indicates the actual number of cycles of the k-th level stress; Indicates the first material parameter; m indicates the second material parameter.

[0055] Figure 4 Is an exemplary schematic diagram of the training acceleration unit provided by the present invention. The model training module further includes a training acceleration unit, as Figure 4 shown, the training acceleration unit includes an initial model determination layer 311, a genetic algorithm update layer 312, an RBF model update layer 313, a first loop layer 314, an update acceleration layer 315, and a second loop layer 316.

[0056] The initial model determination layer determines an initial RBF model based on the initial monitoring point influence factor and the initial neuron weight. The initial monitoring point influence factor and the initial neuron weight are obtained through an initialization model. The initial RBF model refers to an RBF model initialized by the initial monitoring point influence factor and the initial neuron weight.

[0057] The genetic algorithm update layer updates the initial genetic algorithm based on the initial RBF model to obtain a first genetic algorithm and a first monitoring point influence factor.

[0058] The RBF model update layer updates the initial RBF model based on the first monitoring point influence factor to obtain a first RBF model and a first neuron weight.

[0059] The first loop layer is used to repeat the update operations of the genetic algorithm and the RBF model until multiple monitoring point influence factors and multiple neuron weights are obtained;

[0060] The update acceleration layer is used to process the multiple monitoring point influence factors and the multiple neuron weights to obtain an accelerated monitoring point influence factor and an accelerated neuron weight. The multiple monitoring point influence factors and the multiple neuron weights are three, and the obtained accelerated monitoring point influence factor and accelerated neuron weight are:

[0061] ;

[0062] Among them, Indicates the accelerated monitoring point influence factor; Indicates the accelerated neuron weight; , And Respectively represent the third, second, and first monitoring point influence factors, and the fitness values of the third, second, and first monitoring point influence factors decrease in sequence; , And respectively represent the third, second, and first neuron weights, and the objective function values corresponding to the third, second, and first neuron weights decrease in sequence.

[0063] The second loop layer replaces the initial monitoring point influence factor and the initial neuron weight with the acceleration monitoring point influence factor and the acceleration neuron weight respectively, and repeats the update operations of the genetic algorithm and the RBF model until the fitness value of the genetic algorithm converges.

[0064] The update acceleration layer also determines an acceleration objective function value based on the acceleration monitoring point influence factor and the acceleration neuron weight, and compares the acceleration objective function value with the third objective function value; when the acceleration objective function value is greater than or equal to the third objective function value, the second loop layer replaces the initial monitoring point influence factor and the initial neuron weight with the acceleration monitoring point influence factor and the acceleration neuron weight respectively; otherwise, the second loop layer replaces the initial monitoring point influence factor and the initial neuron weight with the third monitoring point influence factor and the third neuron weight respectively. The acceleration objective function value refers to the objective function value calculated through the objective function according to the acceleration monitoring point influence factor and the acceleration neuron weight. The third objective function value refers to the objective function value calculated through the objective function according to the third monitoring point influence factor and the third neuron weight.

[0065] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent monitoring system for vehicle axles, characterized in that: It includes monitoring point selection module, data acquisition module, current damage determination module, remaining life calculation module, model training module and alarm module; The monitoring point selection module screens the monitoring point set through a genetic algorithm to obtain a plurality of monitoring points; The data acquisition module includes a plurality of stress sensors, and the stress sensors are arranged at the plurality of monitoring points to obtain the current stress sequences at the plurality of monitoring points; The current damage determination module is used to process the current stress sequence to obtain the current damage of each monitoring point; The remaining life calculation module processes the current damage and historical damage of the multiple monitoring points through the RBF model to obtain the remaining life of the axle; The model training module is used to train the genetic algorithm and the RBF model. The genetic algorithm and the RBF model are obtained through joint training. The purpose of the joint training is to maximize the objective function. The objective function is: ； in, represents the objective function; , and Respectively represent the first target parameter, the second target parameter and the third target parameter; i represents the monitoring point variable; I represents the total number of monitoring points; represents the remaining lifespan output by the RBF model obtained using the previous round of training; t represents the time variable of the acquired monitoring point; T represents the total number of time points of the data at the acquired monitoring point; represents the actual remaining life at time point t; It represents the predicted remaining life at time point t output by the RBF model obtained in the current round of training; represents the weight of the jth RBF neuron; The alarm module is used to issue an alarm when the remaining life does not meet the safety threshold.

2. The intelligent monitoring system for vehicle axles according to claim 1, characterized in that: The monitoring point selection module includes a construction unit, a screening unit and a circulation unit; The construction unit constructs a plurality of monitoring point selection vectors based on the monitoring point set; the monitoring point selection vectors are used to represent the selection of monitoring points in the monitoring point set; The screening unit calculates the fitness value of each of the monitoring point selection vectors and determines a plurality of target monitoring point selection vectors; the fitness value is related to the sensitivity of the monitoring point to the remaining life; The cyclic unit is used to process the target monitoring point selection vector to obtain multiple new monitoring point selection vectors, and repeat the screening operation until the fitness value converges.

3. The intelligent axle monitoring system according to claim 2, characterized in that: The fitness function for solving the fitness value is: ;in, represents the fitness value; i represents the monitoring point variable; I represents the total number of monitoring points in the monitoring point set; Indicates the selection of the i-th monitoring point; Indicates the position of the i-th monitoring point on the bridge; Indicates the remaining life of the axle; Indicates the sensitivity of the monitoring point to the remaining life.

4. The intelligent monitoring system for vehicle axles according to claim 1, characterized in that: The current damage determination module includes a stress sequence decomposition unit, a cycle number determination unit, an equivalent stress calculation unit and a current damage determination unit; The stress sequence decomposition unit is used to process the current stress sequence to obtain a stress cycle set and cycle characteristics of each stress cycle; the cycle characteristics include stress amplitude, average stress and total number of cycles; The cycle number determination unit determines the current cycle number at each stress level based on the stress amplitude; The equivalent stress calculation unit determines the equivalent stress based on the current number of cycles and the stress level; The current damage determination unit determines the current damage based on the equivalent stress and the current number of cycles.

5. The intelligent monitoring system for vehicle axles according to claim 4, characterized in that: The current stress sequence is processed by a rainflow counting method to obtain the stress cycle set and the cycle characteristics.

6. The intelligent axle monitoring system according to claim 4, characterized in that: The stress range of each stress level is: ; ; ;in, represents the stress of the kth stress level; and They represent the minimum and maximum values ​​of the kth level stress respectively; represents the stress value when the vehicle is unloaded; k represents the stress level variable; Indicates the total number of stress levels; Indicates the stress value when the vehicle is fully loaded; The calculation formula of the equivalent stress is: ;in, represents the equivalent stress of the kth level stress; represents the actual amplitude of the kth stress; represents the average stress of the kth level stress; The current damage is: ;in, Indicates current damage; Indicates the actual number of cycles of the kth stress level; represents the first material parameter; m represents the second material parameter.

7. The intelligent axle monitoring system according to claim 1, characterized in that: The remaining life is: ; Where L represents the predicted remaining life; j represents the hidden layer node variable; J represents the number of hidden layer nodes; represents the weight of the jth RBF neuron; exp represents the exponential function; i represents the monitoring point variable; I represents the total number of monitoring points; Indicates the impact factor of each monitoring point on the remaining life; represents the damage of the i-th monitoring point at the t-th time point; represents the i-th component of the j-th RBF center point; Represents the expansion constant.

8. The intelligent axle monitoring system according to claim 1, characterized in that: The model training module also includes a training acceleration unit, which includes an initial model determination layer, a genetic algorithm update layer, an RBF model update layer, a first circulation layer, an update acceleration layer, and a second circulation layer; The initial model determination layer determines the initial RBF model based on the initial monitoring point influencing factors and initial neuron weights; The genetic algorithm update layer updates the initial genetic algorithm based on the initial RBF model to obtain a first genetic algorithm and a first monitoring point influencing factor; The RBF model updating layer updates the initial RBF model based on the first monitoring point influencing factor to obtain a first RBF model and a first neuron weight; The first circulation layer is used to repeat the updating operation of the genetic algorithm and the RBF model until multiple monitoring point influencing factors and multiple neuron weights are obtained; The update acceleration layer is used to process the multiple monitoring point influence factors and the multiple neuron weights to obtain accelerated monitoring point influence factors and accelerated neuron weights; The second circulation layer replaces the initial monitoring point influence factor and the initial neuron weight with the acceleration monitoring point influence factor and the acceleration neuron weight respectively, and repeats the updating operation of the genetic algorithm and the RBF model until the fitness value of the genetic algorithm converges.

9. The intelligent axle monitoring system according to claim 8, characterized in that: The number of the multiple monitoring point influencing factors and the number of the multiple neuron weights is three, and the obtained acceleration monitoring point influencing factors and acceleration neuron weights are: ;in, represents the impact factor of the accelerated monitoring point; represents the acceleration neuron weight; , and Respectively represent the influencing factors of the third, second and first monitoring points, and the fitness values ​​of the influencing factors of the third, second and first monitoring points decrease in sequence; , and They represent the third, second and first neuron weights respectively, and the objective function values ​​corresponding to the third, second and first neuron weights decrease in sequence.

10. The intelligent axle monitoring system according to claim 9, characterized in that: The update acceleration layer further determines an acceleration objective function value based on the acceleration monitoring point influence factor and the acceleration neuron weight, and compares the acceleration objective function value with a third objective function value; When the acceleration objective function value is greater than or equal to the third objective function value, the second circulation layer replaces the initial monitoring point influence factor and the initial neuron weight respectively with the acceleration monitoring point influence factor and the acceleration neuron weight; Otherwise, the second recurrent layer replaces the initial monitoring point influence factor and the initial neuron weight with the third monitoring point influence factor and the third neuron weight respectively.

Images