Health status prediction method of mechatronic transmission system based on autonomous model selection

Abstract

The present invention relates to the technical field of intelligent operation and maintenance of mechanical equipment, and specifically to a method for predicting the health status of an electromechanical hybrid transmission system based on autonomous model selection. The method comprises: obtaining real-time vibration signals during the operation of the electromechanical hybrid transmission system and extracting effective eigenvalues; determining a starting prediction point for the data based on the effective eigenvalue interval 3#imgabs0#; constructing a model library containing multiple degradation models; when the data reaches the starting prediction point, fitting each model in the model library using a maximum likelihood estimation method, and scoring all models using the KIC criterion to select the optimal model; updating the optimal model parameter distribution using a particle filtering method; and respectively introducing the updated parameter distribution into the optimal model to calculate the health status of the electromechanical hybrid transmission system. The present invention can select the optimal model based on the degradation trend of the electromechanical hybrid transmission system under different conditions and has good prediction performance.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance of mechanical equipment, and more specifically to a method for predicting the health status of an electromechanical composite transmission system based on autonomous model selection. Background Art

[0002] Mechatronic transmission systems are key subsystems of vehicles, so improving their safety and reducing their maintenance costs are current research priorities. Mechatronic transmission systems are subject to harsh operating conditions such as high temperatures and high loads during operation, making them prone to wear, fatigue, and various other failures. Failures in mechatronic transmission systems can lead to delays in work progress and economic losses, while serious accidents can endanger personnel. To promptly detect potential failures in mechatronic transmission systems, predictive maintenance strategies are being adopted to manage and maintain the health of mechatronic transmission systems. Predicting the health status of mechatronic transmission systems is a key technology for health management and maintenance, preventing catastrophic failures, reducing maintenance costs, and effectively improving the safety, reliability, and economy of mechatronic transmission systems.

[0003] Model-based prediction methods are widely used in existing mechatronic transmission system health status prediction. These methods describe the health degradation process of mechatronic transmission systems by establishing mathematical or physical models and updating model parameters based on real-time measured information. In many cases, the health degradation process of mechatronic transmission systems is complex and time-varying, making it difficult to accurately describe it using a specific functional form. Using only a single model cannot accurately fit the complex health degradation process of mechatronic transmission systems. Therefore, to address the problem that simple linear or nonlinear models cannot accurately describe the complex health degradation process of mechatronic transmission systems, scholars have proposed a health status prediction method based on multi-model selection. Although this method improves the shortcomings of the original single-model prediction method, the existing multi-model selection prediction method is affected by the amount of data. When the amount of data is small, it tends to select overly simple models, which reduces the accuracy of the prediction results and affects subsequent maintenance decisions.

[0004] Therefore, how to propose a multi-model selection method that is not affected by the amount of data and apply it to the health status prediction of mechatronic transmission systems to improve the prediction performance is an urgent problem that technical personnel in this field need to solve. Summary of the Invention

[0005] In view of this, the present invention provides a health status prediction method for an electromechanical hybrid transmission system based on autonomous model selection, which can select the optimal model according to the degradation trend of the electromechanical hybrid transmission system under different conditions and has good prediction performance.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for predicting the health status of an electromechanical hybrid transmission system based on autonomous model selection includes the following steps:

[0008] S1. Obtain the real-time vibration signal of the electromechanical hybrid transmission system during operation and extract the effective eigenvalue;

[0009] S2, determine the effective eigenvalue 3 interval, and according to 3 Interval identification of normal and abnormal states of the electromechanical hybrid transmission system, and determination of the starting prediction point of the data;

[0010] S3. Build a model library containing multiple degradation models;

[0011] S4. When the data reaches the starting prediction point, use the maximum likelihood estimation method to fit each model in the model library, and use the KIC criterion to score all models to screen out the optimal model;

[0012] S5. Initialize the parameter distribution of the optimal model, update the parameter distribution using the particle filtering method, and continuously update the particle weights to obtain the updated parameter distribution;

[0013] S6. The updated parameter distribution is respectively introduced into the optimal model to predict the characteristic value of the electromechanical hybrid transmission system at a future time point. When the predicted characteristic value reaches the set failure threshold, the failure time is recorded and the health status of the electromechanical hybrid transmission system is calculated.

[0014] Furthermore, in S1, the extraction formula for extracting effective features from the vibration signal is:

[0015] ;

[0016] in, represents the effective eigenvalue, represents the time domain vibration signal, represents the signal sampling point, .

[0017] Furthermore, in S2, the mean of the effective eigenvalues ​​under normal conditions is calculated and variance , confirm 3 interval , the effective eigenvalue is 3 The data in the interval are normal state data, and the effective eigenvalue is 3 Data outside the interval is abnormal data; the process of determining the starting prediction point of the data includes:

[0018] Initialization time point , increasing the time point in sequence , find the effective eigenvalues More than 3 for the first time Time point of the interval ,Right now ;

[0019] Pick , No. The effective characteristic value at each time point for , when it still satisfies Time, take time The starting point for prediction ;

[0020] Ruodang hour, Dissatisfied , then continue to increase the time point until two consecutive time points and The data values ​​satisfy When, take The starting point of the prediction .

[0021] Furthermore, in S3, the model library contains five degradation models, namely linear model, logarithmic model, double exponential model, Gamma model and S-type function + power exponential function model.

[0022] Furthermore, in S4, the KIC criterion avoids overfitting by penalizing the complexity of the model and selects the model with the lowest KIC value as the optimal model for the current prediction time step; the expression of the KIC criterion is:

[0023] ;

[0024] in, represents the observed random state variable; is the maximum likelihood estimate of the model, including A vector of adjustable parameters; is the probability density; is the log-likelihood function, in the maximum likelihood estimation Get the minimum value at represents the number of observed random state variables; represents the normalized form of the Fisher information matrix, where Depend on composition, is the Fisher information matrix The elements in Depend on Line and Column composition; The calculation formula is:

[0025] ;

[0026] in, and They are middle Row elements and Likelihood estimates for column elements, Indicates the maximum likelihood estimation Calculation The value of .

[0027] Furthermore, S5 includes:

[0028] S51, set the initial prior distribution for all parameters of the selected optimal model, starting from the starting prediction point, and extract particles, each particle represents a set of parameter values;

[0029] S52. Starting from the starting prediction point, obtain the valid feature value at each time point as the observation value;

[0030] S53. For each particle, calculate the importance probability of its generating the observation value at the current time point, and obtain the posterior probability density distribution of the particle swarm;

[0031] S54, calculate the weight of each particle and perform normalization processing; resample from the initial prior distribution, eliminate low-weight particles, and copy high-weight particles;

[0032] The resampled particle swarm forms the parameter distribution of the current time point and serves as the prior distribution of the next time point.

[0033] Furthermore, in S53, the posterior probability density distribution is expressed as:

[0034] ;

[0035] in, To update the time, for The true effective eigenvalue at time , represents the probability, express The sampling time A particle sample, Indicates the total number of sampled particles, represents the Dirac function, express Moment The sample weights, express All samples taken at the moment.

[0036] Furthermore, in S54, the particle weight The calculation formula is as follows:

[0037] ;

[0038] in, To update the time, for The true value of the moment, represents probability; Indicates that it is based on the previous state and the previous observation Generate current state The conditional probability of Indicates that based on the previous observation Generate the previous state The conditional probability of express Momentary status The initial probability distribution of .

[0039] Furthermore, in S6, the remaining life calculation formula of the electromechanical hybrid transmission system is:

[0040] ;

[0041] in, Indicates the time The remaining life value of the electromechanical transmission system is represents the predicted value whose eigenvalue increases with the prediction time, represents the failure threshold, Indicates the increase in cumulative running time, Indicates the All characteristic values ​​obtained at the sampling time.

[0042] Furthermore, S6 also includes: performing statistics on the remaining life of the electromechanical hybrid transmission system predicted by all particles and calculating the mean value and variance , the calculation formulas are:

[0043] ;

[0044] ;

[0045] in, Represents the total number of particles; mean As the final remaining life prediction value; variance Used to quantify uncertainty.

[0046] It can be seen from the above technical solutions that compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention pre-sets a variety of commonly used degradation models in the model library, first through 3 The system identifies normal and abnormal states of mechatronic transmission system operating data within intervals. When an abnormality occurs, it triggers an autonomous degradation model selection process. This improves the prediction of the health of mechatronic transmission systems by overcoming the shortcomings of existing prediction methods that rely solely on a single model. Furthermore, the KIC criterion is used for model selection, selecting the optimal model for data with different degradation trajectories. This balances model complexity and goodness of fit, ensuring better generalization and improving the accuracy of prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0049] Figure 1 A flow chart of the method for predicting the health status of an electromechanical hybrid transmission system based on autonomous model selection provided by the present invention;

[0050] Figure 2 A framework diagram of the method for predicting the health status of an electromechanical hybrid transmission system based on autonomous model selection provided by the present invention;

[0051] Figure 3 The present invention provides a 3- Flowchart of the adaptive condition monitoring algorithm for the interval;

[0052] Figure 4 Schematic diagram showing the performance comparison between the method of the present invention and existing prediction methods. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0054] The embodiment of the present invention discloses a method for predicting the health status of an electromechanical hybrid transmission system based on autonomous model selection, comprising the following steps:

[0055] S1. Obtain the real-time vibration signal of the electromechanical hybrid transmission system during operation and extract the effective eigenvalue;

[0056] S2, determine the effective eigenvalue 3 interval, and according to 3 Interval identification of normal and abnormal states of the electromechanical hybrid transmission system, and determination of the starting prediction point of the data;

[0057] S3. Build a model library containing multiple degradation models;

[0058] S4. When the data reaches the starting prediction point, use the maximum likelihood estimation method to fit each model in the model library, and use the KIC criterion to score all models to screen out the optimal model;

[0059] S5. Initialize the parameter distribution of the optimal model, update the parameter distribution using the particle filtering method, and continuously update the particle weights to obtain the updated parameter distribution;

[0060] S6. The updated parameter distribution is respectively introduced into the optimal model to predict the characteristic value of the electromechanical hybrid transmission system at a future time point. When the predicted characteristic value reaches the set failure threshold, the failure time is recorded and the health status of the electromechanical hybrid transmission system is calculated.

[0061] The following is a further explanation of the above steps. The specific process is as follows: Figure 1-Figure 2 shown.

[0062] S1. Real-time detection of the operating data of the electromechanical hybrid transmission system and extraction of features.

[0063] Specifically, the vibration signal of the electromechanical composite transmission system is collected in real time by sensors , from the vibration signal Extract valid values ​​from , the extraction formula is:

[0064] ;

[0065] in, represents the signal sampling point, .

[0066] For effective eigenvalues Perform maximum-minimum normalization processing, and the normalization formula is:

[0067] .

[0068] S2, use based on 3 The adaptive condition monitoring algorithm of the interval determines the data starting prediction point.

[0069] Specifically, such as Figure 3As shown, calculate the mean of the effective eigenvalues ​​under normal conditions and variance , confirm 3 interval , the effective eigenvalue is 3 The data in the interval are normal state data, and the effective eigenvalue is 3 Data outside the interval is abnormal data; the process of determining the starting prediction point of the data includes:

[0070] S21, initialization time point , increasing the time point in sequence , find the effective eigenvalues More than 3 for the first time Time point of the interval ,Right now ;

[0071] S22, take , No. The effective characteristic value at each time point for , when it still satisfies Time, take time The starting point for prediction ;

[0072] S23. Ruodang hour, Dissatisfied , then continue to increase the time point until two consecutive time points and The data values ​​satisfy When, take The starting point for prediction .

[0073] S3. Build a model library. The model library contains five degradation models: linear model, logarithmic model, double exponential model, Gamma model, and S-type function + power exponential function model. The five fitting model formulas are:

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] ;

[0079] in, is the operating time of the electromechanical hybrid transmission system, are the parameters that need to be fitted.

[0080] S4. Independently select the degradation model for each updated operating data. The specific selection process is as follows:

[0081] S41. Use the maximum likelihood estimation method to fit the model. Since there is no obvious degradation information in the early stage of the electromechanical hybrid transmission system, in order to save monitoring costs, when the data reaches the starting prediction point , it starts autonomous model selection for each updated data, and uses the maximum likelihood estimation method to fit each model in the model library.

[0082] S42. Use the KIC criterion to score all models. The KIC criterion avoids overfitting by penalizing the complexity of the model and selects the model with the lowest KIC value as the optimal model for the current prediction time step. The expression of the KIC criterion is:

[0083] ;

[0084] in, represents the observed random state variable; is the maximum likelihood estimate of the model, including A vector of adjustable parameters; is the probability density; is the log-likelihood function, in the maximum likelihood estimation Get the minimum value at represents the number of observed random state variables; represents the normalized form of the Fisher information matrix, where Depend on composition, is the Fisher information matrix The elements in Depend on Line and Column composition; The calculation formula is:

[0085] ;

[0086] in, and They are middle Row elements and Likelihood estimates for column elements, Indicates the maximum likelihood estimation Calculation The value of .

[0087] Linear model For example, the likelihood function The calculation formula is:

[0088] ;

[0089] in, is the dependent variable, is the eigenvector, are the fitting parameters of the linear model, is the error term, is the variance of the residual. The above formula is used to select models for the operational data. A lower KIC value indicates a better balance between data fitting and computational complexity. Therefore, during the model selection process, the model with the lowest KIC value is selected for subsequent health status prediction.

[0090] S5. Update the model's parameter distribution online using known data, specifically including:

[0091] S51, set the initial prior distribution for all parameters of the selected optimal model, starting from the starting prediction point, and extract particles, each particle represents a set of parameter values.

[0092] S52. Starting from the starting prediction point, obtain the valid feature value at each time point as the observation value;

[0093] S53. For each particle, calculate the importance probability of its generating the observation value at the current time point, and obtain the posterior probability density distribution of the particle swarm; the posterior probability density distribution is expressed as:

[0094] ;

[0095] in, To update the time, for The true effective eigenvalue at time , represents the probability, express The sampling time A particle sample, Indicates the total number of sampled particles, represents the Dirac function, express Moment The sample weights, express All samples taken at the moment.

[0096] S54, calculate the weight of each particle and perform normalization; resample from the initial prior distribution, eliminate low-weight particles, and copy high-weight particles; particle weight The calculation formula is as follows:

[0097] ;

[0098] in, To update the time, for The true value of the moment, represents probability; Indicates that it is based on the previous state and the previous observation Generate current state The conditional probability of Indicates that based on the previous observation Generate the previous state The conditional probability of express Momentary status The initial probability distribution of .

[0099] The weights are normalized, and the normalization formula is:

[0100] ;

[0101] in, express The moment is based on the sampling Particle samples Generate observation data The conditional probability of Indicates the Particles in The weight of the moment; for The normalized form of Indicates the Particles in The weight of the moment (used for recursive update), Indicates the slave state To status The state transition distribution probability of Indicates that based on the current observation and the previous state Generate current state The conditional probability of .

[0102] According to the calculated weights, the samples are resampled, and samples with high weights are reused, while samples with low weights are eliminated.

[0103] Resampling is performed using the inverse cumulative distribution function method, and the resampled particle swarm becomes the posterior distribution of the current time step , and use it as the next time step The prior distribution of .

[0104] S6. Prediction of the health status of electromechanical hybrid transmission systems.

[0105] The updated parameter distribution is respectively introduced into the optimal model to predict the eigenvalues ​​of the electromechanical hybrid transmission system at future time points. As the time step increases, the particle swarm distribution is continuously updated. When the predicted eigenvalue reaches the set failure threshold, the particles stop updating. The time when all particles stop updating, i.e., the failure time, is recorded, and the health status of the electromechanical hybrid transmission system is calculated.

[0106] The remaining service life calculation formula of the electromechanical hybrid transmission system is:

[0107] ;

[0108] in, Indicates the time The remaining life value of the electromechanical transmission system is represents the predicted value whose eigenvalue increases with the prediction time, represents the failure threshold, Indicates the increase in cumulative running time, Indicates the All characteristic values ​​obtained at the sampling time.

[0109] Calculate the mean of the remaining life of the electromechanical transmission system predicted by all particles and variance , the calculation formulas are:

[0110] ;

[0111] ;

[0112] in, Represents the total number of particles; mean As the final remaining life prediction value; variance Used to quantify uncertainty.

[0113] Next, taking rail transit bearing trend prediction as an example, the bearing operation data was collected through the experimental bench to verify the effectiveness of the method of the present invention.

[0114] The test bearing used in this example is an LDK UER204 rolling bearing. The bearing's parameters are shown in Table 1. The experimental platform primarily consists of an AC motor, a frequency converter, a hydraulic loading system, a load indicator, a rotating shaft, a support bearing, a test bearing, and a bearing support. The speed is controlled by the AC motor's frequency converter, and the load is generated by the hydraulic loading system and applied horizontally to the test bearing's bearing support.

[0115] Table 1. LDK UER204 rolling bearing parameters

[0116]

[0117] The vibration signal sampling strategy was set as follows: a sampling frequency of 25.6 kHz, a sampling interval of 1 minute, and a sampling duration of 1.28 seconds (i.e., 32,768 data points). This example employed three different operating conditions, testing five bearings under each condition. Table 2 provides detailed information on the experimental operating conditions and actual operating life of each tested bearing.

[0118] Table 2. Rolling bearing accelerated life test information list

[0119]

[0120] The experiment uses the method proposed in the present invention to select models for bearing operation data in real time and make predictions. The experiment compares the method proposed in the present invention with three other multi-model prediction methods. Method A uses a weight distribution method based on distance metric and combines multiple models for integrated remaining life prediction; Method B uses multivariate linear regression, support vector regression and Gaussian process regression models to generate preliminary remaining life predictions, and finally uses the random forest model to fuse the preliminary remaining life predictions; Method C develops an automatic technical process to trigger the RUL prediction module and select the best fitting model driven by conditional monitoring data adaptively without any manual help. The four methods were compared using different bearings, as shown in Table 3 and Table 3. Figure 4 The comparative effects of the root mean square error and convergence rate of the three types of bearing operating data are listed.

[0121] Table 3 Comparison of root mean square error of four methods

[0122]

[0123] It can be seen from the experimental results that the method of the present invention achieves better prediction results by using the KIC criterion for model selection, and the prediction effects are significantly better than the other three existing methods.

[0124] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0125] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the health status of an electromechanical hybrid transmission system based on autonomous model selection, characterized in that: The following steps are involved: S1. Obtain the real-time vibration signal of the electromechanical hybrid transmission system during operation and extract the effective eigenvalue; S2. Determine the 3σ interval of the effective characteristic value, identify the normal state and abnormal state of the electromechanical hybrid transmission system based on the 3σ interval, and determine the starting prediction point of the data; S3. Build a model library containing multiple degradation models; S4. When the data reaches the starting prediction point, the maximum likelihood estimation method is used to fit each model in the model library, and the KIC criterion is used to score all models to screen out the optimal model. The KIC criterion avoids overfitting by penalizing the complexity of the model and selects the model with the lowest KIC value as the optimal model for the current prediction time step. The expression of the KIC criterion is: Among them, z * represents the observed random state variable; is the maximum likelihood estimate of the model, including W j is a vector of adjustable parameters; p(·) is the probability density; is the log-likelihood function, in the maximum likelihood estimation The minimum value is obtained at z represents the number of observed random state variables; Represents the normalized form of the Fisher information matrix, where R j By R jαβ Composition, R jαβ is the Fisher information matrix R j The elements in R jαβ It consists of α rows and β columns; R jαβ The calculation formula is: Among them, ξ jα and ξ jβ R jαβ The likelihood estimates of the α row elements and β column elements in , Indicates the maximum likelihood estimation Calculate R jαβ The value of S5. Initialize the parameter distribution of the optimal model, update the parameter distribution using the particle filtering method, and continuously update the particle weights to obtain the updated parameter distribution; S6. The updated parameter distribution is respectively introduced into the optimal model to predict the characteristic value of the electromechanical hybrid transmission system at a future time point. When the predicted characteristic value reaches the set failure threshold, the failure time is recorded and the health status of the electromechanical hybrid transmission system is calculated.

2. The method for predicting the health status of an electromechanical hybrid transmission system based on model autonomous selection according to claim 1, characterized in that: In S1, the extraction formula for extracting effective features from vibration signals is: Among them, x RMS Represents the effective eigenvalue, M(n) represents the time domain vibration signal, n represents the signal sampling point, n=1,2,...,N.

3. The method for predicting the health status of a mechatronic transmission system based on model autonomous selection according to claim 1, characterized in that: In S2, the mean μ and variance σ of the effective eigenvalues ​​under normal conditions are calculated, and the 3σ interval [μ-3σ, μ+3σ] is determined. The effective eigenvalues ​​within the 3σ interval are normal state data, and the effective eigenvalues ​​outside the 3σ interval are abnormal state data. The process of determining the starting prediction point of the data includes: Initialize time point g = 0, increase time point g in sequence, and find the effective eigenvalue x RMS The time point g0 that first exceeds the 3σ interval, that is, Take g = g0 + 1, the effective eigenvalue x at the g0 + 1th time point RMS for When it still satisfies When time g0 is taken as the starting prediction point FPT g ; If g=g0+1, Dissatisfied Then continue to increase the time point until two consecutive time points g FPT and g FPT+1 The data values ​​all satisfy |x g When -μ|>3σ, take g FPT is the starting prediction point FPT g .

4. The method for predicting the health status of an electromechanical hybrid transmission system based on model autonomous selection according to claim 1, characterized in that: In S3, the model library contains five degradation models, namely linear model, logarithmic model, double exponential model, Gamma model and S-type function + power exponential function model.

5. The method for predicting the health status of a mechatronic transmission system based on autonomous model selection according to claim 1, characterized in that S5 include: S51. Setting an initial prior distribution for all parameters of the selected optimal model, starting from the starting prediction point, extracting H particles from the initial prior distribution, each particle representing a set of parameter values; S52. Starting from the starting prediction point, obtain the valid feature value at each time point as the observation value; S53. For each particle, calculate the importance probability of the particle generating the observation value at the current time point, and obtain the posterior probability density distribution of the particle swarm; S54, calculate the weight of each particle and perform normalization processing; resample from the initial prior distribution, eliminate low-weight particles, and copy high-weight particles; The resampled particle swarm forms the parameter distribution of the current time point and serves as the prior distribution of the next time point.

6. The method for predicting the health status of an electromechanical hybrid transmission system based on model autonomous selection according to claim 5, characterized in that: In S53, the posterior probability density distribution is expressed as: Among them, m is the update time, y m is the true effective eigenvalue at time m, P(·) represents the probability, represents the i-th particle sample sampled at time m, H represents the total number of sampled particle samples, η(·) represents the Dirac function, represents the weight of the i-th sample at time m, x m Represents all samples sampled at time m.

7. The method for predicting the health status of a mechatronic transmission system based on autonomous model selection according to claim 6, characterized in that: In S54, particle weight The calculation formula is as follows: Among them, m is the update time, y m is the true value at time m, P(·) represents the probability; q(x m |x m-1 ,y m-1 ) represents the state based on the previous state x m-1 and the previous observation y m-1 Generate the current state x m The conditional probability of q(x m-1 |y m-1 ) represents the value based on the previous observation y m-1 Generate the previous state x m-1 The conditional probability of p(x m ) represents the state x at time m m The initial probability distribution of .

8. The method for predicting the health status of an electromechanical hybrid transmission system based on model autonomous selection according to claim 1, characterized in that: In S6, the remaining service life calculation formula of the electromechanical hybrid transmission system is: RUL(t m )=inf{r:Γ(r+t m )≥λ|f}; Among them, RUL(t m ) represents the time t m The remaining life value of the electromechanical hybrid transmission system is Γ(r+t m ) represents the predicted value of the feature value increasing with the prediction time, λ represents the failure threshold, r represents the increase in the cumulative running time, and f represents the tth m All characteristic values ​​obtained at the sampling moment.

9. The method for predicting the health status of a mechatronic transmission system based on autonomous model selection according to claim 8, characterized in that: S6 also includes: performing statistics on the remaining life of the electromechanical hybrid transmission system predicted by all particles, and calculating the mean μ(t m ) and variance σ 2 (t m ), the calculation formulas are: Where H represents the total number of particles; the mean μ(t m ) as the final remaining life prediction value; variance σ 2 (t m ) is used to quantify uncertainty.

Images