Degradation state estimation method for electromechanical equipment components based on uncertain particle filtering

Abstract

The application belongs to the technical field of mechanical and electrical equipment fault diagnosis, and particularly relates to a mechanical and electrical equipment component degradation state estimation method based on uncertain particle filtering, which comprises the following steps: S1, system modeling is performed on the mechanical and electrical equipment component to determine the state equation and observation equation of the mechanical and electrical equipment component; S2, according to the initial state of the mechanical and electrical equipment component, uncertain particles at the first time are obtained, and an initial uncertain distribution function is determined; S3, according to the state equation of the mechanical and electrical equipment component, uncertain particles at the second time are calculated, and a priori uncertain distribution of the state to be estimated of the mechanical and electrical equipment component is obtained; S4, real-time estimation of the degradation state of the mechanical and electrical equipment component; and S5, degradation state estimation of the mechanical and electrical equipment component at the next time is performed. The application can accurately estimate the state of the mechanical and electrical equipment component through real-time monitoring data, has high sensitivity, can quickly predict the real state of the object, and has a good application prospect in the field of fault diagnosis.

Description

Technical Field

[0001] The present invention belongs to the technical field of electromechanical equipment fault diagnosis, and in particular relates to a method for estimating the degradation state of electromechanical equipment components based on uncertain particle filtering. Background Art

[0002] Fault prediction involves predicting the degradation trend of a device based on its current health state, ultimately estimating the device's remaining useful life (RUL) using predefined thresholds. For large, complex electromechanical equipment, early warning of equipment anomalies can prevent sudden failures, significantly improving equipment maintainability and availability, and reducing overall lifecycle costs. Predictive models may not be able to monitor the current health state of components within electromechanical equipment in a timely manner. For example, battery degradation may change during charging and discharging, but this cannot be directly and accurately measured within the component. Bearing wear also occurs continuously during operation, but it is impossible to stop the equipment and measure the wear precisely in real time. To enhance the effectiveness of fault prediction models, it is crucial to estimate the real-time health state as accurately and timely as possible. In electromechanical equipment, the health state can be indirectly represented by one or a set of monitored variables, and a mapping relationship can be established between the real-time health state and observed values.

[0003] In fault prediction scenarios, the health state is time-varying, and the mapping function may not be consistent throughout the degradation process. Bayesian filtering has emerged as an optimal solution, offering the advantage of being able to use new observations to update the prior distribution, bringing it closer to the true distribution. The particle filter discretizes the distribution into particles, avoiding the computationally intensive infinite integral calculations required in Bayesian filtering. The particle filter model includes system modeling and uncertainty management. System modeling consists of the state transition equation and the observation equation. Uncertainty management is the core of the particle filter, which uses observations to reduce the uncertainty of unknown parameters. Uncertainty management involves four steps: particle initialization, prediction step, update step, and resampling.

[0004] However, particle filtering also has some shortcomings. One of them is the lag between the estimated value of the unknown parameter and the change in the true value. When the degradation pattern changes and the true value of the unknown variable may change suddenly, the particle filtering method requires a sufficient number of observations to approximate the changed value. In extreme cases where the true value changes frequently and the number of observations is limited, the particle filtering method may not be able to accurately estimate the real-time health status of the equipment, resulting in unsatisfactory performance of RUL prediction. The core reason for this shortcoming is that the Bayesian method has good stability but poor flexibility to new observations. Therefore, it is necessary to propose a sensitive and highly accurate real-time estimation method for the degradation status of electromechanical equipment components. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a method for estimating the degradation state of electromechanical equipment components based on uncertain particle filtering. It accurately estimates the state of electromechanical equipment components through real-time monitoring data. It has high sensitivity and can quickly predict the true state of the observed object when the observation data is limited. It has considerable application prospects in the field of fault diagnosis.

[0006] To achieve the above objectives, the present invention discloses the following technical solutions:

[0007] A method for estimating degradation states of electromechanical equipment components based on uncertain particle filtering, comprising:

[0008] S1: Conduct system modeling of electromechanical equipment components and determine the state equations and observation equations of the electromechanical equipment components;

[0009] The system modeling of the electromechanical equipment components to be estimated is carried out, and the state equations G(ξ t ) and the observation equation H(ξ t );

[0010] S2: According to the initial state of the electromechanical equipment components, the uncertain particles at the first moment are obtained and the initial uncertainty distribution function is determined;

[0011] The state of the electromechanical equipment component to be estimated is set as the uncertain variable ξ t , using uncertain particle value x t Fit the initial distribution function and determine the initial uncertain distribution function as:

[0012]

[0013] Among them, Ψ(ξ t ) is the initial uncertainty distribution function; ξ t is an uncertain variable, which represents the state of the electromechanical equipment components at the first moment t; t ) is the particle distribution function; Φ i (x t ) is the distribution function of the i-th uncertain variable; x t The value of the uncertain particle at the first moment; Take the value for the i-th uncertain particle; is the reliability of the i-th uncertain particle; i is the number of the uncertain particle; n is the total number of uncertain particles; δ is the uncertainty distribution solution function; t is the first moment; min is the minimum value function;

[0014] S3: Calculate the uncertain particles at the second moment according to the state equation of the electromechanical equipment component to obtain the prior uncertainty distribution of the state to be estimated of the electromechanical equipment component;

[0015] Set the error function of the state equation of the electromechanical equipment component in step S1 to q t , obtain the initial uncertainty distribution function in step S2, calculate the prior distribution function of the uncertain variable at the second time t+1, further calculate the uncertain particles at the second time, and obtain the prior uncertainty distribution of the state to be estimated of the electromechanical equipment component:

[0016]

[0017] Among them, - (ξ t+1 ) is the prior uncertainty distribution of the state to be estimated of the electromechanical equipment components; ξ t+1 is the state of the electromechanical equipment component at the second moment t+1, x t+1 is the value of the uncertain particle at the second moment; t+1 is the second moment;

[0018] S4: Real-time estimation of degradation state of electromechanical equipment components: Using the particle-based uncertain Bayesian theorem, the credibility of the uncertain particles at the second moment is calculated to obtain the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment components, and complete the real-time estimation of the degradation state of the electromechanical equipment components; the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment components is:

[0019]

[0020] Among them, + (ξ t+1 ) is the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment component; η is the posterior uncertainty distribution coefficient; F′(y s |x t+1 ) is the likelihood function of uncertain particles; Ψ -′ (x t+1 ) is the derivative of the prior uncertainty distribution; y s is the observation value of the sth electromechanical equipment component; s is the observation value number of the electromechanical equipment component; m is the total number of observation values ​​of the electromechanical equipment component;

[0021] Posterior uncertainty distribution Ψ of the estimated state of electromechanical equipment components + (ξ t+1 ) is the estimated state value of the key equipment of the electromechanical equipment at the second moment;

[0022] S5: Estimating the degradation state of the electromechanical equipment components at the next moment: Use linear interpolation as the resampling method to generate new uncertain particles, set a resampling threshold based on the number of valid uncertain particles and a resampling threshold based on the confidence interval, and resample the uncertain particles when the resampling threshold is exceeded; repeat steps S3 and S4 with the obtained new uncertain particles to complete the degradation state estimation of the electromechanical equipment components at the next moment.

[0023] Preferably, the state equation of the electromechanical equipment component in step S1 is specifically:

[0024] The system modeling is carried out according to the state of the electromechanical equipment components, and the state equation of the electromechanical equipment components is obtained as follows:

[0025] ξ t+1 =G(ξ t )+q t ;

[0026] Among them, G(ξ t ) is the uncertainty variable ξ t The equation of q t is the error function of the state equation at the first moment t;

[0027] The state equation of the electromechanical equipment component is used to represent the process of transition between degradation states, that is, the uncertain variable ξ that characterizes the degradation state at the first moment t t and the uncertain variable ξ that characterizes the degradation state at the second moment t+1 t+1 The relationship between the first moment t and the second moment t+1 is predicted by the degradation state.

[0028] Preferably, the observation equation of the electromechanical equipment component in step S1 is specifically:

[0029] y t+1 =H(ξ t+1 )+r t+1 ;

[0030] Among them, y t+1 is the observed value of the state of the electromechanical equipment component to be estimated at the second time t+1; H(ξ t+1 ) is the uncertain variable ξ at the second moment t+1 t+1 The observation equation of r t+1 is the observation error at the second moment t+1;

[0031] The observation equation of the electromechanical equipment component reflects the relationship between the degradation state and the observation value. After obtaining the observation value at the second time t+1, the posterior uncertainty distribution corresponding to the prior uncertainty distribution obtained by the state equation is obtained.

[0032] Preferably, in step S2, the state of the electromechanical equipment component to be estimated is set as the uncertain variable ξ t , using uncertain particle value x t Fit the initial distribution function, specifically:

[0033] S21: Set the state of the electromechanical equipment component to be estimated as an uncertain variable, and determine the uncertain variable to be estimated ξ t The initial distribution of is subject to normal uncertainty distribution;

[0034] S22: Uncertain particles including their values and its reliability According to the uncertain variable ξ t The distribution of uncertain distribution samples is generated The uncertain distribution sample Arrange in ascending order, and there are no duplicate values; is an uncertain variable ξ t The value is less than or equal to The minimum and maximum values ​​of uncertain particles are written as and

[0035] S23: Using uncertain particles to fit the initial distribution function Ψ(ξ t ); uncertain variables represents the i-th uncertain particle, and The relationship between them is:

[0036]

[0037] in, is the mean of the uncertain variable;

[0038] At the first moment t, the distribution function corresponding to each uncertain variable is:

[0039]

[0040] Among them, Φ n (x t ) is the nth uncertain variable The distribution function of

[0041] For any set uncertain particle parameter ε, the discretized uncertain particles satisfy the following inequality:

[0042]

[0043] in, is an uncertain variable ξ t Uncertain measure of is an uncertain variable The uncertainty measure of ; max is the maximum value function; x min is the lower limit of the value of uncertain particles; x max is the upper limit of the value of uncertain particles; ε is the parameter of uncertain particles;

[0044] Therefore, ξ t The uncertainty distribution is used within any precision The uncertainty distribution is approximated; when When , the initial distribution function of particle is obtained.

[0045] Preferably, in step S3, the error function of the state equation of the electromechanical equipment component in step S1 is set to q t , obtain the initial uncertainty distribution function in step S2, calculate the prior distribution function of the uncertain variable at the second time t+1, and further calculate the uncertain particles at the second time, specifically:

[0046] S31: Determine the error function q of the state equation of the electromechanical equipment component t , which obeys the normal uncertainty distribution, is

[0047] S32: Determine the uncertain variable ξ at the second time t+1 t+1 The prior distribution function Ψ - (x t+1 ); According to the state equation of the electromechanical equipment components, we have ξ t+1 =G(ξ t )+q t Established, suppose that about the uncertain variable ξ t The equation G(ξ t ) has a distribution function of Φ G , then the error function q t The distribution function is Φ q , then its overall distribution function is:

[0048]

[0049] Among them, - is the uncertain variable ξ t+1 Prior distribution function of G For the uncertain variable ξ t The equation G(ξ t ) distribution function; Φ q is the error function q t The distribution function of G(x t ) is the uncertain variable ξ t The value of Q t is the error function q t The value of sup is the function x t+1 =G(x t )+Q t The supremum of

[0050] In the error function q t When the value of is determined, the state equation of the electromechanical equipment components becomes:

[0051] x t+1 =G(xt )+Q t ;

[0052] S33: Solve the above equation to obtain the uncertain variable ξ at the second moment t+1 t+1 According to step S32, for each determined x t+1 , if and only if Φ G (G(x t ))=Φ q (Q t ), G(x t ) and Q t Has a unique and definite value; the equations are established as follows:

[0053]

[0054] In the above formula, the error function q t The distribution function is Φ q It is known that due to the uncertain variable ξ t The equation G(ξ t ) is about the uncertain variable ξ t function, the two have the same reliability, that is, they satisfy Φ G (G(x t ))=Ψ(x t ); Solving the above equations, we can get G(x t ) and Q t The only definite value, the result of the prior uncertainty distribution function is:

[0055] Ψ - (x t+1 )=Φ G (G(x t ))=Φ q (Q t );

[0056] Among them, - (x t+1 ) Uncertain variable ξ t+1 The prior uncertainty distribution function results of ;

[0057] At this time, the prior uncertainty distribution function Ψ - It is the uncertain variable ξ at the second moment t+1 t+1 Corresponding reliability;

[0058] S34: According to the calculation process from step S31 to step S33, find the value of each uncertain variable at the second time t+1 Different values The corresponding reliability Among them, the value of the uncertain particle pass Calculated, the corresponding reliability Solving the above equations, we can obtain:

[0059] For the first moment t, when the number of uncertain particles meets the requirement, the uncertain variable ξ t+1 Smaller than the uncertain particle x t+1 The reliability of the state of the electromechanical equipment components to be estimated is replaced by uncertain particles, which can obtain the prior uncertainty distribution Ψ - (ξ t+1 ).

[0060] Preferably, in step S4, the particle-based uncertain Bayesian theorem is used to calculate the credibility of the uncertain particles at the second moment, obtain the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment component, and complete the real-time estimation of the degradation state of the electromechanical equipment component, specifically:

[0061] S41: Determine the uncertain variable ξ t+1 The prior uncertainty distribution function result Ψ - (x t+1 ); According to step S34, the uncertain variable ξ can be obtained t+1 The prior uncertainty distribution function of is Ψ - (x t+1 )for:

[0062]

[0063] S42: Determine the likelihood function of the uncertain particles; for the uncertain particle filter, the likelihood function is: Among them, y s Represents the observed values, y1, y2, ..., y m ;x t+1 is the uncertain variable ξt at time t+1 +1 Here, the distribution of the observed data is known, that is, F(y s |x t+1 ) is known, and the specific distribution is set according to the actual observation data;

[0064] S43: Get the uncertain variable ξ t+1 The posterior uncertainty distribution function Ψ + (x t+1 ); Combining the likelihood function and the prior uncertainty distribution, the posterior uncertainty distribution function is:

[0065]

[0066] Among them, + (x t+1 ) is the posterior uncertainty distribution function result; η is the posterior uncertainty distribution coefficient;

[0067] The posterior uncertainty distribution coefficient is:

[0068]

[0069] According to the discretization method of uncertain particle filtering, the posterior uncertainty distribution coefficient is converted into:

[0070]

[0071] Thus, the posterior uncertainty distribution Ψ of the degradation state to be estimated of the electromechanical equipment components is determined + (ξ t+1 ).

[0072] Preferably, in step S5, linear interpolation is used as a resampling method to generate new uncertain particles, a resampling threshold based on the number of valid uncertain particles and a resampling threshold based on the confidence interval are set, and resampling of uncertain particles is performed when the resampling threshold is exceeded, specifically:

[0073] S51: Determine the resampling method and generate new uncertain particles; from step S4 and step S5, it can be concluded that the uncertain particle filtering method uses the state equation ξ of the electromechanical device component t+1 =G(ξ t )+q t To determine the value of the uncertain particle at the second moment and calculate its prior confidence, the discretized uncertain Bayesian theorem is used to update the posterior confidence of the uncertain particle at the second moment, and obtain the information of the uncertain particle at the second moment; determine the two adjacent uncertain particles and If the corresponding value This indicates that the phenomenon of particle swallowing has occurred. Too many uncertain particles swallowed will greatly affect the effect of state estimation. In the resampling process, the linear interpolation method is used to generate new uncertain particles. When two adjacent uncertain variables Its uncertain particle value is The corresponding reliability is Then the linear interpolation of these two uncertain particles is:

[0074]

[0075] Among them, a′ t+1 is the uncertainty particle reliability; x′ t+1 Take values ​​for uncertain particles;

[0076] Set any uncertain particle value x′ t+1 , and its corresponding reliability a′ t+1 Calculate from the above formula and get a new uncertain particle (x′t+1 ,a′ t+1 );

[0077] S52: Determine a resampling threshold based on the effective number of uncertain particles and a resampling threshold based on the confidence interval; the resampling threshold based on the effective number of uncertain particles is set according to the effective number of uncertain particles in the uncertain particle filtering method as follows:

[0078]

[0079] in, is the number of valid uncertain particles at the second time t+1; N is the total number of initialized uncertain particles;

[0080] Set a threshold for the number of effective uncertain particles. If the calculated number of effective uncertain particles is lower than the set threshold, resampling of uncertain particles is required.

[0081] The resampling threshold based on the confidence interval; after calculating the posterior uncertainty distribution, the expected value e of the uncertain variable can be obtained t+1 ; The value range of uncertain particles should include the set confidence interval:

[0082]

[0083] in, is the minimum value of the uncertain particle at the second moment t+1; is the maximum value of the uncertain particle at the second moment t+1; e t+1 is the expected value of the uncertain variable; is the upper limit of the expected value of the uncertain variable; is the lower bound of the expected value of the uncertain variable;

[0084] By setting the confidence interval, the upper and lower limits of the confidence interval can be recorded as and When the value of the uncertain particle in step S4 does not satisfy the above formula, resampling is required.

[0085] Preferably, in step S51, a linear interpolation method is used to generate new uncertain particles, specifically:

[0086] S511: Obtain the posterior uncertainty distribution of uncertain particles;

[0087] S512: According to the method for obtaining the resampling threshold in step S51, determine whether the particles of the posterior uncertainty distribution meet the resampling criteria. If so, proceed to steps S513-S516; if not, do not proceed to steps S513-S516.

[0088] S513: Determine the number of resampled uncertain particles to ensure that the number of resampled uncertain particles is the same as the original number of uncertain particles;

[0089] S514: Determine the uncertain variable ξ that needs to be resampled t+1 The value range of

[0090] S515: Randomly generate new uncertain particle values; within the set value range, randomly generate m uncertain particles in the order of small to large, that is,

[0091] S516: Calculate the reliability of the new uncertain particle and obtain the reliability corresponding to the new uncertain particle Generate new uncertain particles after resampling.

[0092] Compared with the prior art, the present invention has the following beneficial effects:

[0093] (1) The method for estimating the degradation state of electromechanical equipment components based on uncertain particle filtering provided by the present invention performs system modeling on the components of the electromechanical equipment to be estimated, regards the unknown state as an uncertain variable, and gives an uncertain particleization method to obtain the initial uncertain distribution function after particleization, generates new uncertain particles, and solves the corresponding reliability of the new particles to obtain the prior distribution of the uncertain variable; finally, the uncertain Bayesian theorem after particleization is used to update the reliability of the uncertain particles at the second moment, and obtain the posterior uncertainty distribution of the uncertain variable, thereby realizing the degradation state estimation of the electromechanical equipment components.

[0094] (2) The present invention takes into account the particle swallowing phenomenon that exists in the particle filtering process. In order to reduce the impact of particle degradation on the state estimation prediction results of electromechanical equipment components, an uncertain particle resampling method is proposed. According to the set resampling threshold and method, the particles are resampled to ensure the accuracy of the state estimation of electromechanical equipment components.

[0095] (3) Compared with traditional methods, this method has higher sensitivity and can accurately estimate the degradation state of electromechanical equipment when there is less observation data. It also shows high sensitivity during the observation of outliers. This method has a wide range of applications and can be used for fault diagnosis and condition monitoring of complex equipment such as motor devices and transmission devices in electromechanical equipment, where the degradation state cannot be directly observed. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 This is a control block diagram of a method for estimating degradation states of electromechanical equipment components based on uncertain particle filtering according to the present invention;

[0097] Figure 2 is the image of the distribution function of the uncertain variable of the present invention;

[0098] Figure 3 is the state equation error function q t Schematic diagram of the distribution curve;

[0099] Figure 4 is a graph showing the estimated degradation state of a lithium-ion battery according to the present invention;

[0100] Figure 5 A graph showing the estimated degradation state of lithium-ion batteries using comparative methods. DETAILED DESCRIPTION

[0101] The exemplary embodiments, features, and aspects of the present invention will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0102] The present invention analyzes lithium-ion batteries. For large, complex electromechanical devices, power supply units are crucial for providing electrical power. Therefore, power supply performance is crucial, as power supply failure often signals failure of the electromechanical device. However, due to the complex structure of electromechanical devices, it is impossible to directly monitor the status of the batteries in the power supply units, making accurate fault diagnosis impossible. The proposed uncertain particle filtering method can observe the output data of electromechanical devices, thereby predicting battery degradation and enabling accurate fault diagnosis.

[0103] The present invention provides a method for estimating the degradation state of electromechanical equipment components based on uncertain particle filtering, such as Figure 1 As shown, a system model is performed on the electromechanical equipment component, the state equation and observation equation of the electromechanical equipment component are determined, and based on the initial state of the electromechanical equipment component, the uncertain particles at the first moment are obtained, and the initial uncertain distribution function is determined. Based on the state equation of the electromechanical equipment component, the uncertain particles at the second moment are calculated to obtain the prior uncertainty distribution of the state to be estimated of the electromechanical equipment component, the degradation state of the electromechanical equipment component is estimated in real time, and the degradation state of the electromechanical equipment component at the next moment is estimated. The embodiment of the present invention takes the degradation process of lithium-ion batteries as the specific research object, and its specific steps include:

[0104] Step S1: Perform system modeling on the lithium-ion battery and determine the state equation and observation equation of the lithium-ion battery.

[0105] A system model is constructed for lithium-ion batteries, a key component of electromechanical equipment, to determine their state equations and observation equations. For lithium-ion batteries, the remaining capacity is often used to reflect the battery's degradation state. The battery degradation process can be divided into two stages: the battery capacity degrades slowly at the beginning and then rapidly before failure. A double exponential model is used to describe the battery capacity degradation process, and its state equation is as follows:

[0106] ξ t+1 =G(ξ t )+q t ;

[0107] Among them, ξ t+1 Represents the remaining capacity of the lithium-ion battery at the first moment t+1, which is an uncertain variable; G(ξ t ) is the uncertainty variable ξ t The equation of , in Example 1, is confirmed to be a double exponential degradation model; q t is the error function of the state equation at the first moment t, which is an uncertain variable and obeys the distribution

[0108] The state equation of lithium-ion batteries is used in the field of fault prediction to represent the process of transition between degradation states, that is, the uncertain variable ξ that characterizes the degradation state at the first moment t t and the uncertain variable ξ that characterizes the degradation state at the second moment t+1 t+1 The relationship between the first moment t and the second moment t+1 is predicted by the degradation state.

[0109] The system model is built based on the state of the lithium-ion battery, and the observation equation of the lithium-ion battery is obtained as follows:

[0110] y t+1 =H(ξ t+1 )+r t+1 ;

[0111] Among them, y t+1 is the observed value monitored at the second moment t+1; H(ξ t+1 ) is the remaining capacity of the lithium-ion battery at the second moment t+1 ξ t+1 The observation equation of r t+1 is the observation error at the second moment t+1.

[0112] This observation error is caused by the accuracy of the sensor. In the observation equation, the form and parameters of the observation equation are known. Therefore, after obtaining the observation value at time t+1, the posterior uncertainty distribution corresponding to the prior uncertainty distribution obtained by the state equation is calculated. In electromechanical equipment, the observation value at a certain moment is only related to the degradation state at that moment. Usually, the degradation amount of the internal components of electromechanical equipment cannot be directly observed, and the only observation data that can be obtained are easily observable external variables, such as the current and voltage output by the motor equipment, the operation trajectory of the power equipment, etc. The observation equation of the lithium-ion battery reflects the relationship between the degradation state of the lithium-ion battery and the monitored observation value; after obtaining the observation value at the second moment t+1, the posterior uncertainty distribution corresponding to the prior uncertainty distribution obtained by the state equation can be calculated.

[0113] Step S2: According to the initial state of the lithium-ion battery, the uncertain particles at the first moment are obtained and the initial uncertainty distribution function is determined.

[0114] like Figure 2 The image of the distribution function of the uncertain variable of the present invention is shown; the state to be estimated of the lithium-ion battery is set as the uncertain variable ξ t , using uncertain particle value x t Fit the initial distribution function, specifically:

[0115] Step S21: The state of the lithium-ion battery to be estimated is set as the uncertain variable ξ t , determine the uncertain variable ξ t The initial distribution of is subject to normal uncertainty distribution.

[0116] Step S22: Uncertain particles including their values and its reliability According to the uncertain variable ξ t The distribution of uncertain particles is generated Uncertain particles Arrange in ascending order, and there are no duplicate values; is an uncertain variable ξ t The value is less than or equal to The minimum and maximum values ​​of uncertain particles are written as and

[0117] Step S23: Use uncertain particles to fit the initial distribution function Ψ(ξ t ); uncertain variables represents the i-th uncertain particle, and The relationship between them is:

[0118]

[0119] in, is the mean of the uncertain variable.

[0120] At the first moment t, the distribution function corresponding to each uncertain variable is:

[0121]

[0122] Among them, Φ n (x t ) is the nth uncertain variable The distribution function of .

[0123] For any set uncertain particle parameter ε, the discretized uncertain particles satisfy the following inequality:

[0124]

[0125] in, is an uncertain variable ξ t Uncertain measure of is an uncertain variable The uncertainty measure of ; max is the maximum value function; x min is the lower limit of the value of uncertain particles; x max is the upper limit of the value of uncertain particles; ε is the parameter of uncertain particles.

[0126] Therefore, ξ t The uncertainty distribution is used within any precision The uncertainty distribution is approximated; when When , the initial uncertainty distribution function can be determined as:

[0127]

[0128] Among them, Ψ(ξ t ) is the initial uncertainty distribution function; ξ t is an uncertain variable, indicating the state of the lithium-ion battery at the first moment t; t ) is the particle distribution function; Φ i (x t ) is the distribution function of the i-th uncertain variable; x t The value of the uncertain particle at the first moment; Take the value for the i-th uncertain particle; is the reliability of the i-th uncertain particle; i is the number of the uncertain particle; n is the total number of uncertain particles; δ is the distribution solution function of the uncertain particles; t is the first moment; min is the minimum value function.

[0129] Step S3: Calculate the uncertain particles at the second moment according to the state equation of the lithium-ion battery to obtain a priori uncertainty distribution of the state to be estimated of the lithium-ion battery.

[0130] Step S31: Set the error function of the lithium-ion battery state equation in step S1 to q t , according to the initial uncertainty distribution function in step S2, calculate the prior distribution function of the uncertain variable at the second moment t+1, and further calculate the uncertain particles at the second moment; determine the error function q of the lithium-ion battery state equation t , which obeys the normal uncertainty distribution, is Its distribution curve is as follows Figure 3 shown.

[0131] Step S32: Determine the uncertain variable ξ at the second time t+1 t+1 The prior distribution function Ψ - (x t+1 ); According to the state equation of lithium-ion batteries, there is ξ t+1 =G(ξ t )+q t Established, suppose that about the uncertain variable ξ t The equation G(ξ t ) has a distribution function of Φ G , then the error function q t The distribution function is Φ q , then its overall distribution function is:

[0132]

[0133] Among them, - is the uncertain variable ξ t+1 Prior distribution function of Φ G For the uncertain variable ξ t The equation G(ξ t ) distribution function; Φ q is the error function q t The distribution function of G(x t ) is the uncertain variable ξ t The value of Q t is the error function q t The value of sup is the function x t+1 =G(x t )+Q t The supremum of .

[0134] Therefore, in the error function q t When the value of is determined, the state equation of the lithium-ion battery becomes as follows:

[0135] x t+1 =G(xt )+Q t ;

[0136] Step S33: Solve the above equation to obtain the uncertain variable ξ at the second time t+1 t+1 According to step S32, for each determined x t+1 , if and only if Φ G (G(x t ))=Φ q (Q t ), G(x t ) and Q t Has a unique and definite value; the equations are established as follows:

[0137]

[0138] In the above formula, the error function q t The distribution function is Φ q It is known that due to the uncertain variable ξ t The equation G(ξ t ) is about the uncertain variable ξ t function, the two have the same reliability, that is, they satisfy Φ G (G(x t ))=Ψ(x t ); Solving the above equations, we can get G(x t ) and Q t The only definite value, the result of the prior uncertainty distribution function is:

[0139] Ψ - (x t+1 )=Φ G (G(x t ))=Φ q (Q t );

[0140] Among them, - (x t+1 ) Uncertain variable ξ t+1 The prior uncertainty distribution function results.

[0141] At this time, the prior uncertainty distribution function Ψ - It is the uncertain variable ξ at the second moment t+1 t+1 The corresponding reliability.

[0142] Step S34: According to the above calculation process, find each uncertain variable at the second time t+1 Different values The corresponding reliability Among them, the value of the uncertain particle pass Calculated, the corresponding reliability It is obtained by solving the above equations.

[0143] For the first moment t, when the number of uncertain particles meets the requirement, the uncertain variable ξ t+1 Smaller than the uncertain particle x t+1 The confidence of the lithium-ion battery is replaced by uncertain particles, that is, the prior uncertainty distribution Ψ of the estimated state of the lithium-ion battery can be obtained. - (ξ t+1 )for:

[0144]

[0145] Among them, - (ξ t+1 ) is the prior uncertainty distribution of the lithium-ion battery state to be estimated; ξ t+1 is the state of the lithium-ion battery at the second moment t+1, x t+1 The value of the uncertain particle at the second moment; t+1 is the second moment.

[0146] Step S4: Real-time estimation of the degradation state of the lithium-ion battery: Using the particle-based uncertain Bayesian theorem, the credibility of the uncertain particles at the second moment is calculated to obtain the posterior uncertainty distribution of the lithium-ion battery state to be estimated, and the real-time estimation of the degradation state of the lithium-ion battery is completed.

[0147] Step S41: Determine the uncertain variable ξ t+1 The prior uncertainty distribution function result Ψ - (x t+1 ); According to step S34, the uncertain variable ξ can be obtained t+1 The prior uncertainty distribution function of is Ψ - (x t+1 )for:

[0148]

[0149] Step S42: Determine the likelihood function of the uncertain particles; for the uncertain particle filter, the likelihood function is: Among them, y s Represents the observed values, y1, y2, ..., y m ;x t+1 is the uncertain variable ξt at time t+1 +1 Here, the distribution of the observed data is known, that is, F(y s |x t+1 ) is known, and the specific distribution is set according to the actual observed data.

[0150] Step S43: Obtain the uncertain variable ξt+1 The posterior uncertainty distribution function Ψ + (x t+1 ); Combining the likelihood function and the prior uncertainty distribution, the posterior uncertainty distribution function is:

[0151]

[0152] Among them, + (x t+1 ) is the posterior uncertainty distribution function result; η is the posterior uncertainty distribution coefficient.

[0153] The posterior uncertainty distribution coefficient is:

[0154]

[0155] According to the discretization method of uncertain particle filtering, the above posterior uncertainty distribution function is sorted out, and the posterior uncertainty distribution coefficient is converted into:

[0156]

[0157] Further analysis can determine the posterior uncertainty distribution Ψ of the estimated degradation state of the lithium-ion battery. + (ξ t+1 )for:

[0158]

[0159] Among them, + (ξ t+1 ) is the posterior uncertainty distribution of the estimated state of the lithium-ion battery; η is the posterior uncertainty distribution coefficient; F′(y s |x t+1 ) is the likelihood function of uncertain particles; Ψ -′ (x t+1 ) is the derivative of the prior uncertainty distribution; y s is the sth observation value; s is the observation value number; m is the total number of observation values.

[0160] At this time, the posterior uncertainty distribution Ψ of the estimated state of the lithium-ion battery + (ξ t+1 ) is the estimated state of the lithium-ion battery at the second moment.

[0161] Step S5: Estimating the degradation state of the lithium-ion battery at the next moment.

[0162] To avoid particle swallowing, new uncertain particles are generated using linear interpolation. A resampling threshold based on the number of valid uncertain particles and a resampling threshold based on the confidence interval are set. When the resampling threshold is exceeded, the uncertain particles are resampled. The specific methods for obtaining the resampling threshold based on the number of valid uncertain particles and the resampling threshold based on the confidence interval are as follows:

[0163] Step S51: Determine the resampling method of the uncertain particles; the uncertain particle filtering method derived from step S4 uses the state equation ξ of the lithium-ion battery t+1 =G(ξ t )+q t To predict the value of the uncertain particle and calculate its prior confidence; for the mth uncertain particle, the uncertain variable and The relationship between them is as follows:

[0164]

[0165] Since there is an error function q in the state equation of lithium-ion batteries t , when the uncertain variable The value is less than When the value of , the following equation holds:

[0166]

[0167] The above formula indicates that two adjacent particles exchange positions. This phenomenon is called particle engulfment. If particle engulfment occurs, the two particles Unable to form ξ t+1 Uncertain distribution. In this way, these two particles need to be discarded. Due to the existence of particle swallowing, the number of effective uncertain particles in the degradation state prediction will continue to decrease, and the accuracy of the prediction results will decrease. At this time, it is necessary to resample the particles; in the resampling process, the linear interpolation method is used to generate new uncertain particles, specifically:

[0168] Step S511: Through steps S3 and S4, the posterior uncertainty distribution of the uncertain particles is derived.

[0169] Step S512: Determine whether the particles of the posterior uncertainty distribution meet the resampling criteria by using the resampling threshold acquisition method. If so, proceed to steps S513-S516; if not, proceed to the following steps.

[0170] Step S513: Determine the number of resampled uncertain particles to ensure that the number of resampled uncertain particles is the same as the original number of uncertain particles.

[0171] Step S514: Determine the uncertain variable ξ that needs to be resampledt+1 The value range of .

[0172] Step S515: Randomly generate new uncertain particle values; within the set value range, randomly generate 500 particles in the order of small to large, that is,

[0173] Step S516: Calculate the reliability of the new uncertain particle and obtain the reliability corresponding to the new uncertain particle Generate new uncertain particles after resampling.

[0174] When two adjacent uncertain variables Its uncertain particle value is The corresponding reliability is Then the linear interpolation of these two uncertain particles is:

[0175]

[0176] Among them, a′ t+1 is the uncertainty particle reliability; x′ t+1 Value for uncertain particles.

[0177] Set any uncertain particle value x′ t+1 , and its corresponding reliability a′ t+1 Calculate from the above formula and get a new uncertain particle (x′ t+1 ,a′ t+1 ).

[0178] Step S52: Determine a resampling threshold based on the number of effective uncertain particles and a resampling threshold based on the confidence interval to limit the computational burden and the decrease in computational efficiency caused by overly frequent resampling.

[0179] Too frequent resampling imposes a significant computational burden and reduces computational efficiency. Therefore, resampling is only necessary when particles have degraded to a certain degree. Therefore, an appropriate threshold is crucial to ensuring the effectiveness and efficiency of the uncertain particle filtering method. The present invention provides two resampling thresholds: one based on the number of valid particles and one based on a confidence interval.

[0180] The resampling threshold based on the effective number of uncertain particles is set according to the effective number of uncertain particles in the uncertain particle filtering method:

[0181]

[0182] in, is the number of valid uncertain particles at the second moment t+1; N is the total number of initialized uncertain particles; in this example, the total number of initial particles is 500. Indicates the reliability of the particle. Set the threshold of the number of valid particles If the calculated number of valid particles is less than 100, particle resampling is required.

[0183] Set the threshold of the effective uncertain particle number. If the calculated effective uncertain particle number is lower than the set threshold, resampling of the uncertain particles is required.

[0184] Resampling threshold based on confidence interval; after calculating the posterior uncertainty distribution, the expected value e of the uncertain variable can be obtained t+1 ; The value range of uncertain particles should include the set confidence interval:

[0185]

[0186] in, is the minimum value of the uncertain particle at the second moment t+1; is the maximum value of the uncertain particle at the second moment t+1; e t+1 is the expected value of the uncertain variable; is the upper limit of the expected value of the uncertain variable; is the lower bound of the expected value of the uncertain variable.

[0187] By setting the confidence interval, the upper and lower limits of the confidence interval can be recorded as and The confidence interval size is usually selected as 90% or 95%. When the value of the uncertain particle obtained by the posterior uncertainty distribution in step S4 does not satisfy the above formula, resampling is required. After the above steps, the state of the lithium-ion battery is estimated in real time, and the state ξ of the lithium-ion battery at the second time t+1 is obtained according to step S4. t+1 The estimated results.

[0188] like Figure 4 The figure shows a lithium-ion battery degradation state estimation curve according to the present invention; the dashed line represents the actual remaining capacity change of the lithium-ion battery, and the solid line represents the estimated capacity of the lithium-ion battery obtained using the uncertain particle filtering method. It can be seen that the estimated remaining capacity of the lithium-ion battery using the uncertain particle filtering method is accurate and consistent with the actual situation.

[0189] like Figure 5The figure shows the lithium-ion battery degradation state estimation results using a comparative method. The comparative method used is the currently more commonly used particle filtering method. Again, the dashed line represents the actual remaining capacity change of the lithium-ion battery, while the solid line represents the estimated capacity of the lithium-ion battery obtained using the particle filtering method. Comparing the two methods shows that the uncertain particle filtering method proposed in this invention offers significant advantages in real-time degradation state estimation, being more sensitive and accurate than existing traditional methods.

[0190] Embodiment 2 of the present invention can also be applied to bearings in electromechanical equipment. Bearings are crucial components in electromechanical equipment. Their primary function is to support rotating mechanical bodies and reduce the mechanical load friction coefficient during transmission. Their precision, performance, lifespan, and reliability play a decisive role in the overall lifespan of the equipment. During use, bearings often experience various phenomena, such as fatigue, wear, corrosion, deformation, and cracks, leading to bearing degradation. When degradation accumulates to a certain level, it leads to bearing failure, and thus to failure of the electromechanical equipment. These failures do not occur suddenly, but rather accumulate slowly over time with equipment use. Therefore, timely observation and assessment of bearing degradation is crucial for lifespan prediction of electromechanical equipment. However, due to the unique nature of bearing structures, it is impossible to monitor wear or crack length in a timely manner, making direct observation of degradation impossible. Therefore, an uncertain particle filtering method can be used to determine internal bearing failure by observing observable variables such as bearing speed. In this embodiment, bearing track wear can be used as the state variable to be estimated, treating it as an uncertain variable. Bearing speed is treated as observed data. Using a common physical model of bearing wear failure as the state degradation equation, an observation equation is established through historical data fitting or physical modeling. This equation represents the relationship between bearing track wear and bearing speed. Steps S1 through S5 are then implemented, using an uncertain particle filter to estimate bearing wear in real time, enabling real-time bearing fault diagnosis.

[0191] The present invention has the following beneficial effects: The present invention implements system modeling for lithium-ion batteries, treating the unknown state as an uncertain variable with an uncertain distribution at the first moment. A method for particleization is then proposed to obtain an initial uncertainty distribution function after particleization, generate new uncertain particles, and solve for the corresponding beliefs of the new particles to obtain a prior distribution of the uncertain variable. Finally, the uncertain Bayesian theorem after particleization is used to update the beliefs of the uncertain particles at the moment of particleization, obtaining a posterior uncertainty distribution of the uncertain variable, thereby achieving state estimation of the lithium-ion battery. To reduce the impact of particle degradation on lithium-ion battery state estimation and prediction results, the present invention proposes an uncertain particle resampling method. Particles are resampled according to a set resampling threshold and method to ensure the accuracy of lithium-ion battery state estimation. Compared to traditional methods, this method has higher sensitivity and can accurately estimate the degradation state of lithium-ion batteries when there is less observation data. It also exhibits high sensitivity during outlier observation. This method has a wide range of applications and can also be used for fault diagnosis and condition monitoring of complex equipment such as bearings, motors, and transmissions in electromechanical equipment, where the degradation state cannot be directly observed, while ensuring the accuracy of the results.

[0192] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering, characterized in that: It includes: S1: Conduct system modeling of electromechanical equipment components and determine the state equations and observation equations of the electromechanical equipment components; The system modeling of the electromechanical equipment components to be estimated is carried out, and the state equations G(ξ t ) and the observation equation H(ξ t ); S2: According to the initial state of the electromechanical equipment components, the uncertain particles at the first moment are obtained and the initial uncertainty distribution function is determined; The state of the electromechanical equipment component to be estimated is set as the uncertain variable ξ t , using uncertain particle value x t Fit the initial distribution function and determine the initial uncertain distribution function as: Among them, Ψ(ξ t ) is the initial uncertainty distribution function; ξ t is an uncertain variable, which represents the state of the electromechanical equipment components at the first moment t; t ) is the particle distribution function; Φ i (x t ) is the distribution function of the i-th uncertain variable; x t Take the value of the uncertain particle at the first moment; Take the value for the i-th uncertain particle; is the reliability of the i-th uncertain particle; i is the number of the uncertain particle; n is the total number of uncertain particles; δ is the uncertainty distribution solution function; t is the first moment; min is the minimum value function; S3: Calculate the uncertain particles at the second moment according to the state equation of the electromechanical equipment component to obtain the prior uncertainty distribution of the state to be estimated of the electromechanical equipment component; Set the error function of the state equation of the electromechanical equipment component in step S1 to q t , obtain the initial uncertainty distribution function in step S2, calculate the prior distribution function of the uncertain variable at the second time t+1, further calculate the uncertain particles at the second time, and obtain the prior uncertainty distribution of the state to be estimated of the electromechanical equipment component: Among them, - (ξ t+1 ) is the prior uncertainty distribution of the state to be estimated of the electromechanical equipment components; ξ t+1 is the state of the electromechanical equipment component at the second moment t+1, x t+1 is the value of the uncertain particle at the second moment; t+1 is the second moment; S4: Real-time estimation of degradation state of electromechanical equipment components: Using the particle-based uncertain Bayesian theorem, the credibility of the uncertain particles at the second moment is calculated to obtain the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment components, and complete the real-time estimation of the degradation state of the electromechanical equipment components; the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment components is: Among them, + (ξ t+1 ) is the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment component; η is the posterior uncertainty distribution coefficient; F′(y s |x t+1 ) is the likelihood function of uncertain particles; Ψ - ′(x t+1 ) is the derivative of the prior uncertainty distribution; y s is the observation value of the sth electromechanical equipment component; s is the observation value number of the electromechanical equipment component; m is the total number of observation values ​​of the electromechanical equipment component; Posterior uncertainty distribution Ψ of the estimated state of electromechanical equipment components + (ξ t+1 ) is the estimated state value of the key equipment of the electromechanical equipment at the second moment; S5: Estimating the degradation state of the electromechanical equipment components at the next moment: Use linear interpolation as the resampling method to generate new uncertain particles, set a resampling threshold based on the number of valid uncertain particles and a resampling threshold based on the confidence interval, and resample the uncertain particles when the resampling threshold is exceeded; repeat steps S3 and S4 with the obtained new uncertain particles to complete the degradation state estimation of the electromechanical equipment components at the next moment.

2. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 1, characterized in that: The state equation of the electromechanical equipment component in step S1 is specifically: The system modeling is carried out according to the state of the electromechanical equipment components, and the state equation of the electromechanical equipment components is obtained as follows: x t+1 =G(ξ t )+q t ; Among them, G(ξ t ) is the uncertainty variable ξ t The equation of q t is the error function of the state equation at the first moment t; The state equation of the electromechanical equipment component is used to represent the process of transition between degradation states, that is, the uncertain variable ξ that characterizes the degradation state at the first moment t t and the uncertain variable ξ that characterizes the degradation state at the second moment t+1 t+1 The relationship between the first moment t and the second moment t+1 is predicted by the degradation state.

3. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 1, characterized in that: The observation equation of the electromechanical equipment component in step S1 is specifically: y t+1 =H(ξ t+1 )+r t+1 ; Among them, y t+1 is the observed value of the state of the electromechanical equipment component to be estimated at the second time t+1; H(ξ t+1 ) is the uncertain variable ξ at the second moment t+1 t+1 The observation equation of r t+1 is the observation error at the second moment t+1; The observation equation of the electromechanical equipment component reflects the relationship between the degradation state and the observation value. After obtaining the observation value at the second time t+1, the posterior uncertainty distribution corresponding to the prior uncertainty distribution obtained by the state equation is obtained.

4. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 1, characterized in that: In step S2, the state of the electromechanical device component to be estimated is set as the uncertain variable ξ t , using uncertain particle value x t Fit the initial distribution function, specifically: S21: Set the state of the electromechanical equipment component to be estimated as an uncertain variable, and determine the uncertain variable to be estimated ξ t The initial distribution of is subject to normal uncertainty distribution; S22: Uncertain particles including their values and its reliability According to the uncertain variable ξ t The distribution of uncertain distribution samples is generated The uncertain distribution sample Arrange in ascending order, and there are no duplicate values; is an uncertain variable ξ t The value is less than or equal to The minimum and maximum values ​​of uncertain particles are written as and S23: Using uncertain particles to fit the initial distribution function Ψ(ξ t ); uncertain variables represents the i-th uncertain particle, and The relationship between them is: in, is the mean of the uncertain variable; At the first moment t, the distribution function corresponding to each uncertain variable is: Among them, Φ n (x t ) is the nth uncertain variable The distribution function of For any set uncertain particle parameter ε, the discretized uncertain particles satisfy the following inequality: in, is an uncertain variable ξ t Uncertain measure of is an uncertain variable The uncertainty measure of ; max is the maximum value function; x min is the lower limit of the value of uncertain particles; x max is the upper limit of the value of uncertain particles; ε is the parameter of uncertain particles; Therefore, ξ t The uncertainty distribution is used within any precision The uncertainty distribution is approximated; when When , the initial distribution function of particle is obtained.

5. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 2, characterized in that: In step S3, the error function of the state equation of the electromechanical equipment component in step S1 is set to q t , obtain the initial uncertainty distribution function in step S2, calculate the prior distribution function of the uncertain variable at the second moment t+1, and further calculate the uncertain particles at the second moment, specifically: S31: Determine the error function q of the state equation of the electromechanical equipment component t , which obeys the normal uncertainty distribution, is S32: Determine the uncertain variable ξ at the second time t+1 t+1 The prior distribution function Ψ - (x t+1 ); According to the state equation of the electromechanical equipment components, we have ξ t+1 =G(ξ t )+q t Established, suppose that about the uncertain variable ξ t The equation G(ξ t ) has a distribution function of Φ G , then the error function q t The distribution function is Φ q , then its overall distribution function is: Among them, - is the uncertain variable ξ t+1 Prior distribution function of Φ G For the uncertain variable ξ t The equation G(ξ t ) distribution function; Φ q is the error function q t The distribution function of G(x t ) is the uncertain variable ξ t The value of Q t is the error function q t The value of sup is the function x t+1 =G(x t )+Q t The supremum of In the error function q t When the value of is determined, the state equation of the electromechanical equipment components becomes: x t+1 =G(x t )+Q t ; S33: Solve the above equation to obtain the uncertain variable ξ at the second moment t+1 t+1 According to step S32, for each determined x t+1 , if and only if Φ G (G(x t ))=Φ q (Q t ), G(x t ) and Q t Has a unique and definite value; the equations are established as follows: In the above formula, the error function q t The distribution function is Φ q It is known that due to the uncertain variable ξ t The equation G(ξ t ) is about the uncertain variable ξ t function, the two have the same reliability, that is, they satisfy Φ G (G(x t ))=Ψ(x t ); Solving the above equations, we can get G(x t ) and Q t The only definite value, the result of the prior uncertainty distribution function is: P - (x t+1 )=Φ G (G(x t ))=Φ q (Q t ); Among them, - (x t+1 ) Uncertain variable ξ t+1 The prior uncertainty distribution function results of ; At this time, the prior uncertainty distribution function Ψ - It is the uncertain variable ξ at the second moment t+1 t+1 Corresponding reliability; S34: According to the calculation process from step S31 to step S33, find the value of each uncertain variable at the second time t+1 Different values The corresponding reliability Among them, the value of the uncertain particle pass Calculated, the corresponding reliability Solving the above equations, we can obtain: For the first moment t, when the number of uncertain particles meets the requirement, the uncertain variable ξ t+1 Smaller than the uncertain particle x t+1 The reliability of the state of the electromechanical equipment components to be estimated is replaced by uncertain particles, which can obtain the prior uncertainty distribution Ψ - (ξ t+1 ).

6. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 1, characterized in that: In step S4, the particle-based uncertain Bayesian theorem is used to calculate the credibility of the uncertain particles at the second moment, obtain the posterior uncertainty distribution of the state to be estimated of the electromechanical equipment component, and complete the real-time estimation of the degradation state of the electromechanical equipment component. Specifically, S41: Determine the uncertain variable ξ t+1 The prior uncertainty distribution function result Ψ - (x t+1 ); According to step S34, the uncertain variable ξ can be obtained t+1 The prior uncertainty distribution function of is Ψ - (x t+1 )for: S42: Determine the likelihood function of the uncertain particles; for the uncertain particle filter, the likelihood function is: Among them, y s Represents the observed values, y1, y2, ..., y m ;x t+1 is the uncertain variable ξ at time t+1 t+1 Here, the distribution of the observed data is known, that is, F(y s |x t+1 ) is known, and the specific distribution is set according to the actual observation data; S43: Get the uncertain variable ξ t+1 The posterior uncertainty distribution function Ψ + (x t+1 ); Combining the likelihood function and the prior uncertainty distribution, the posterior uncertainty distribution function is: Among them, + (x t+1 ) is the posterior uncertainty distribution function result; η is the posterior uncertainty distribution coefficient; The posterior uncertainty distribution coefficient is: According to the discretization method of uncertain particle filtering, the posterior uncertainty distribution coefficient is converted into: Thus, the posterior uncertainty distribution Ψ of the degradation state to be estimated of the electromechanical equipment components is determined + (ξ t+1 ).

7. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 1, characterized in that: In step S5, linear interpolation is used as the resampling method to generate new uncertain particles. A resampling threshold based on the number of valid uncertain particles and a resampling threshold based on the confidence interval are set. When the resampling threshold is exceeded, the uncertain particles are resampled. Specifically, S51: Determine the resampling method and generate new uncertain particles; from step S4 and step S5, it can be concluded that the uncertain particle filtering method uses the state equation ξ of the electromechanical equipment component t+1 =G(ξ t )+q t To determine the value of the uncertain particle at the second moment and calculate its prior confidence, the discretized uncertain Bayesian theorem is used to update the posterior confidence of the uncertain particle at the second moment, and obtain the information of the uncertain particle at the second moment; determine the two adjacent uncertain particles and If the corresponding value This indicates that the phenomenon of particle swallowing has occurred. Too many uncertain particles swallowed will greatly affect the effect of state estimation. In the resampling process, the linear interpolation method is used to generate new uncertain particles. When two adjacent uncertain variables Its uncertain particle value is The corresponding reliability is Then the linear interpolation of these two uncertain particles is: Among them, a′ t+1 is the uncertainty particle reliability; x′ t+1 Take values ​​for uncertain particles; Set any uncertain particle value x′ t+1 , and its corresponding reliability a′ t+1 Calculate from the above formula and get a new uncertain particle (x′ t+1 ,a′ t+1 ); S52: Determine a resampling threshold based on the effective number of uncertain particles and a resampling threshold based on the confidence interval; the resampling threshold based on the effective number of uncertain particles is set according to the effective number of uncertain particles in the uncertain particle filtering method as follows: in, is the number of valid uncertain particles at the second time t+1; N is the total number of initialized uncertain particles; Set a threshold for the number of effective uncertain particles. If the calculated number of effective uncertain particles is lower than the set threshold, resampling of uncertain particles is required. The resampling threshold based on the confidence interval; after calculating the posterior uncertainty distribution, the expected value e of the uncertain variable can be obtained t+1 ; The value range of uncertain particles should include the set confidence interval: in, is the minimum value of the uncertain particle at the second moment t+1; is the maximum value of the uncertain particle at the second moment t+1; e t+1 is the expected value of the uncertain variable; is the upper limit of the expected value of the uncertain variable; is the lower bound of the expected value of the uncertain variable; By setting the confidence interval, the upper and lower limits of the confidence interval can be recorded as and When the value of the uncertain particle in step S4 does not satisfy the above formula, resampling is required.

8. The method for estimating degradation state of electromechanical equipment components based on uncertain particle filtering according to claim 7, characterized in that: In step S51, a linear interpolation method is used to generate new uncertain particles, specifically: S511: Obtain the posterior uncertainty distribution of uncertain particles; S512: According to the method for obtaining the resampling threshold in step S51, determine whether the particles of the posterior uncertainty distribution meet the resampling criteria. If so, proceed to steps S513-S516; if not, do not proceed to steps S513-S516. S513: Determine the number of resampled uncertain particles to ensure that the number of resampled uncertain particles is the same as the original number of uncertain particles; S514: Determine the uncertain variable ξ that needs to be resampled t+1 The value range of S515: Randomly generate new uncertain particle values; within the set value range, randomly generate m uncertain particles in the order of small to large, that is, S516: Calculate the reliability of the new uncertain particle and obtain the reliability corresponding to the new uncertain particle Generate new uncertain particles after resampling.

Images