Method and system for predicting residual service life of switching power supply

Abstract

The invention relates to the technical field of switch power supply life prediction, in particular to a prediction method and system for the residual service life of a switch power supply, and the method comprises the steps: obtaining a vibration signal of a target switch power supply during opening, and carrying out the noise reduction of the vibration signal based on variational mode decomposition, so as to obtain a noise reduction vibration signal; and performing windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and calculating a short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence. Through combined use of the first life prediction model based on the short-time vibration energy value sequence and the second life prediction model based on the gray model, the influence of different factors on the power supply life, such as the working environment, the load change and the short-circuit current, can be comprehensively considered, and the prediction result is more comprehensive and accurate through combination of the two models; and the limitation of a single factor is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of switch power supply life prediction, and in particular to a method and system for predicting the remaining service life of a switch power supply. Background Art

[0002] A switching power supply (SPS) is a power supply device that generates a stable output voltage by switching the input voltage through a switching element (such as a transistor). Switching power supplies are widely used in various electronic devices, including computers, televisions, communication equipment, etc. Its main function is to convert the input power (usually AC) into the required stable DC voltage.

[0003] Traditional methods often lack effective noise suppression methods. Noise and interference in vibration signals may affect the quality of data, resulting in inaccurate analysis and prediction results. Traditional methods usually rely on preset models or rules. When dealing with complex changes in equipment working conditions, it is difficult to timely reflect the changes in the actual working status of the equipment, resulting in insufficient flexibility in prediction. Traditional methods may only consider the impact of a single factor (such as temperature, load, etc.) on the life of the equipment, while ignoring other possible important factors, such as vibration signals, changes in the working environment, etc. Such methods may not be able to fully reflect the health status of the equipment. Traditional methods usually rely on regular inspections and maintenance and cannot provide real-time and accurate remaining life predictions, which may result in the equipment failing to detect potential problems in time before failure occurs, resulting in unexpected downtime and losses. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a method and system for predicting the remaining service life of a switching power supply.

[0005] The technical solution adopted to solve the above technical problem is: a method for predicting the remaining service life of a switching power supply, comprising:

[0006] Acquire a vibration signal of a target switching power supply when the switch is opened, and perform noise reduction on the vibration signal based on variational mode decomposition to obtain a noise-reduced vibration signal;

[0007] Performing windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and calculating the short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence;

[0008] Extracting the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and performing life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply;

[0009] Acquire a first remaining service life sequence under different short-circuit currents based on the first service life prediction model, and construct a second service life prediction model corresponding to the first remaining service life sequence based on a grey model;

[0010] The short-circuit current is predicted for life based on the second life prediction model to obtain a second remaining service life of the target switching power supply.

[0011] Preferably, denoising the vibration signal based on variational mode decomposition to obtain a denoised vibration signal comprises:

[0012] Decomposing the vibration signal into a plurality of IMF components based on variational mode decomposition;

[0013] Calculate the mutual correlation coefficient between the multiple IMF components and the vibration signal, wherein the calculation formula of the mutual correlation coefficient is as follows:

[0014]

[0015] Among them, ρ k represents the mutual correlation coefficient between the IMF component and the vibration signal, N represents the number of sampling points, and f n Indicates vibration signal, represents the mean value of the vibration signal, u kn represents the kth IMF component, represents the mean value of the kth IMF component;

[0016] Calculate the threshold of the mutual correlation coefficient, wherein the calculation formula of the threshold of the mutual correlation coefficient is as follows:

[0017]

[0018] Among them, η represents the threshold of the cross-correlation coefficient, max(ρ k ) represents the maximum value of the mutual correlation coefficient between the IMF component and the vibration signal;

[0019] Comparing the mutual correlation coefficient with the standard deviation of the mutual correlation coefficient, if the mutual correlation coefficient is greater than the threshold of the mutual correlation coefficient, taking the IMF component corresponding to the mutual correlation coefficient as the effective IMF component;

[0020] The effective IMF components are reconstructed based on singular spectrum analysis to obtain a noise-reduced vibration signal.

[0021] Preferably, the expression of the windowing process is as follows:

[0022]

[0023] Wherein, ω(n) represents the windowing process of the noise reduction vibration signal, L represents the window length of the windowing window, and n represents the total length of the noise reduction vibration signal;

[0024] The calculation formula of the short-time vibration energy value is as follows:

[0025]

[0026] Wherein, E(t) represents the short-time vibration energy value of the short-time vibration signal, [t1, t2] represents the time interval of the short-time vibration signal, and x(t) represents the short-time vibration signal.

[0027] Preferably, extracting the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence includes:

[0028] Presetting a low threshold value, and identifying whether the short-time vibration signal corresponding to the short-time vibration energy value enters a starting state based on the low threshold value, and if the short-time vibration energy value is greater than the low threshold value, it indicates that it is in the starting state;

[0029] A high threshold is pre-set, and based on the high threshold, it is identified whether the short-time vibration signal corresponding to the short-time vibration energy value enters a termination state, and if the short-time vibration energy value is greater than the high threshold, it indicates that it is in a termination state;

[0030] When the short-time vibration energy value is greater than the lower threshold, the time corresponding to the short-time vibration energy value at this time is taken as the initial point;

[0031] The short-time vibration energy value sequence is traversed after the initial point, and when the short-time vibration energy value is greater than the upper threshold, the time corresponding to the short-time vibration energy value at this time is taken as the end point, and the action time is obtained based on the initial point and the end point;

[0032] The above operation is repeatedly performed until the short-time vibration energy value sequence is traversed to obtain the action time sequence.

[0033] Preferably, performing life prediction on the action time series based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply includes:

[0034] The action time degradation characteristic quantity of the target switching power supply is defined, wherein the expression of the action time degradation characteristic quantity is as follows:

[0035] X(t)=a+μt+σ B B(t);

[0036] Among them, X(t) represents the degradation characteristic of action time, a represents the initial value of action time, μ represents the drift coefficient, σ B represents the diffusion coefficient, B(t) represents the standard Wiener process, and t represents the action time series;

[0037] Define the relative increment of the action time degradation characteristic quantity, wherein the relative increment expression is as follows:

[0038] ΔX(t i )=a+μt i +σ B ΔB(t i );

[0039] Among them, ΔX(t i ) represents the relative increment of the action time degradation characteristic quantity, t i represents the i-th action time in the action time series, ΔB(t i ) represents the relative increment of the standard Wiener process;

[0040] Calculating the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity based on maximum likelihood estimation;

[0041] Construct a remaining life probability density function of the target switching power supply, substitute the drift parameter and diffusion coefficient of the relative increment of the action time degradation characteristic quantity into the remaining life probability density function, and use the time corresponding to the maximum value of the remaining life probability density function as the first remaining service life of the target switching power supply, wherein the expression of the remaining life probability density function is as follows:

[0042]

[0043] Among them, f(t|μ,σ B ,h) represents the remaining life probability density function, and h represents the failure threshold.

[0044] Preferably, calculating the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity based on maximum likelihood estimation includes:

[0045] Calculate the likelihood function of the relative increment of the action time degradation feature quantity, wherein the expression of the likelihood function is as follows:

[0046]

[0047] Among them, L(μ,σ B 2 ) represents the likelihood function of the relative increment of the action time degradation feature, I represents the total number of action times in the action time series, g(ΔX(t i ); μ, σ B 2 ) represents the probability density function of the relative increment of the action time degradation feature quantity;

[0048] The likelihood function of the relative increment of the action time degradation feature quantity is solved to obtain the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity, wherein the calculation formula of the drift parameter and the diffusion coefficient is as follows:

[0049]

[0050] in, represents the estimated value of the drift parameter, represents an estimate of the diffusion coefficient.

[0051] Preferably, constructing a second life prediction model corresponding to the first remaining useful life sequence based on a grey model includes:

[0052] The first remaining useful life sequence is accumulated to obtain a life accumulation sequence, wherein the accumulation formula is as follows:

[0053]

[0054] Among them, x (1) (k) represents the kth lifetime accumulation in the lifetime accumulation sequence, x (0) (k) represents the kth first remaining useful life in the first remaining useful life sequence;

[0055] A grey differential equation is established for the life accumulation sequence, wherein the expression of the grey differential equation is as follows:

[0056] x (0) (k)+αz (1) (k) = β;

[0057] in, α represents the development coefficient of the grey differential equation, β represents the grey action of the grey differential equation;

[0058] The grey differential equation is transposed to obtain a grey prediction model, wherein the grey prediction model is expressed as follows:

[0059]

[0060] Preferably, constructing a second life prediction model corresponding to the first remaining useful life sequence based on a grey model further includes:

[0061] Calculating the development coefficient and grey action amount of the grey prediction model based on the least square method;

[0062] Substitute the development coefficient and the gray action into the gray prediction model, and solve the gray prediction model after substitution to obtain the solution of the gray prediction model after substitution, wherein the expression of the solution of the gray prediction model after substitution is as follows:

[0063]

[0064] in, represents the solution of the grey prediction model after substitution;

[0065] The solution of the grey prediction model after substitution is restored, and a second life prediction model is obtained based on the restored grey prediction model.

[0066] Preferably, the expression of the second life prediction model is as follows:

[0067]

[0068] in, represents the second life prediction model, represents the restored grey prediction model, and x (0) (1) represents the first remaining useful life in the first remaining useful life sequence.

[0069] The technical solution adopted to solve the above technical problem is: a system for predicting the remaining service life of a switching power supply, which is applicable to the method for predicting the remaining service life of a switching power supply, comprising:

[0070] A signal noise reduction unit, the signal noise reduction unit is used to obtain a vibration signal of the target switching power supply when the switch is open, and to reduce the noise of the vibration signal based on variational mode decomposition to obtain a noise-reduced vibration signal;

[0071] an energy calculation unit, the energy calculation unit being used to perform windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and calculate the short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence;

[0072] a first prediction unit, the first prediction unit being used to extract the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and to perform life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply;

[0073] A model building unit, the model building unit is used to extract the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and perform life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply;

[0074] A second prediction unit, wherein the second prediction unit is used to perform life prediction on the short-circuit current based on the second life prediction model to obtain a second remaining service life of the target switching power supply.

[0075] The beneficial effects of the present invention are as follows:

[0076] (1) The present invention performs noise reduction processing on the vibration signal through VMD, which effectively removes the noise and interference in the signal and retains the useful signal features, which can improve the quality of the signal and provide more accurate input for subsequent analysis and prediction. In addition, through windowing processing and calculation of short-time vibration energy values, the time and frequency characteristics in the vibration signal can be refined. This refined processing helps to capture subtle changes in the equipment degradation process and further improve the accuracy of life prediction.

[0077] (2) The present invention utilizes a double threshold method to extract the action time, and can automatically detect changes in the working state of the switching power supply. This method has good robustness under different working conditions, can adapt to different working modes of the equipment, and helps to dynamically reflect the use status of the equipment. By modeling the first remaining service life sequence under different short-circuit currents, the influence of current fluctuations on the life of the switching power supply can be considered, which makes the prediction model more adaptable and can perform accurate remaining service life prediction under different loads and working conditions.

[0078] (3) The present invention can comprehensively consider the influence of different factors on the life of the power supply, such as working environment, load change, short-circuit current, etc., by jointly using a first life prediction model based on a short-time vibration energy value sequence and a second life prediction model based on a gray model. The combination of the two makes the prediction result more comprehensive and accurate, avoiding the limitation of a single factor. In addition, by predicting the first and second remaining service life of the target switching power supply, the health status and potential failure risk of the switching power supply can be judged in advance, which helps to perform maintenance or replacement in advance, avoid production downtime or losses due to equipment failure, and improve the reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 A schematic flow chart of the steps of an overall method in an embodiment of the present invention;

[0080] Figure 2 A schematic diagram of the system architecture of an overall system in an embodiment of the present invention.

[0081] Figure numerals: 1. signal noise reduction unit; 2. energy calculation unit; 3. first prediction unit; 4. model building unit; 5. second prediction unit. DETAILED DESCRIPTION

[0082] Embodiment 1, as Figure 1 As shown, the present invention proposes a method for predicting the remaining service life of a switching power supply, comprising:

[0083] S1. Obtain a vibration signal of a target switching power supply when the switch is opened, and perform noise reduction on the vibration signal based on variational mode decomposition to obtain a noise-reduced vibration signal;

[0084] S2. performing windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and calculating the short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence;

[0085] S3, extracting the action time of the short-time vibration energy value sequence based on the double threshold method to obtain an action time sequence, and performing life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply;

[0086] S4. Acquire a first remaining service life sequence under different short-circuit currents based on the first service life prediction model, and construct a second service life prediction model corresponding to the first remaining service life sequence based on a grey model;

[0087] S5. Perform lifetime prediction on the short-circuit current based on the second lifetime prediction model to obtain a second remaining lifetime of the target switching power supply.

[0088] In the present invention, the switching power supply is a power converter that converts voltage through high-speed switching elements (such as transistors). It is more efficient than traditional linear power supplies and is widely used in various electronic devices. In the power supply system, disconnection refers to the process of actively or passively disconnecting the circuit when an electrical device (such as a switching power supply) fails or performs a control operation. This process will induce physical vibration inside or outside the power supply, forming a vibration signal with a certain frequency and amplitude. Windowing is a technology in signal processing, which is usually used to divide the signal into several smaller time periods (windows) so as to further analyze the signal in each window, and to window the vibration signal after noise reduction to obtain a series of short-time signal sequences. The short-time vibration energy value refers to the energy of the signal in each time window. In a physical sense, the energy of the signal is proportional to its vibration. The square of the amplitude is proportional to the power supply. By calculating the energy of each short-time signal, the working status or potential failure of the power supply within the period can be evaluated. The double threshold method is a signal processing method used to extract specific events or time points from continuous signals. The double threshold method is used to extract the "action time" in the short-time vibration energy sequence - that is, the time point when the equipment undergoes a significant change or failure. The action time sequence is a time point sequence extracted from the short-time vibration energy sequence by the double threshold method. Each action time represents a significant change in the switching power supply at a specific time point. The short-circuit current is the current flowing through the circuit when a short-circuit fault occurs in the power supply. Based on different short-circuit current values, the prediction model generates a first remaining service life sequence, each value corresponds to the remaining service life of the switching power supply under a specific short-circuit current. The grey model (Grey Model l, GM) is a mathematical model used to predict, analyze and optimize uncertain systems. It is particularly suitable for situations where information is incomplete or data is scarce. The grey model is used to further construct a second life prediction model to make a more accurate remaining life prediction based on the first remaining service life sequence.

[0089] Embodiment 2, a method for predicting the remaining service life of a switching power supply proposed by the present invention, compared with embodiment 1, this embodiment further includes: denoising the vibration signal based on variational mode decomposition to obtain a denoised vibration signal, including:

[0090] A1. Decompose the vibration signal into multiple IMF components based on variational mode decomposition;

[0091] A2. Calculate the mutual correlation coefficients between the multiple IMF components and the vibration signal, wherein the calculation formula of the mutual correlation coefficient is as follows:

[0092]

[0093] Among them, ρ k represents the mutual correlation coefficient between the IMF component and the vibration signal, N represents the number of sampling points, and fn Indicates vibration signal, represents the mean value of the vibration signal, u kn represents the kth IMF component, represents the mean value of the kth IMF component;

[0094] A3. Calculate the threshold of the mutual correlation coefficient, wherein the calculation formula of the threshold of the mutual correlation coefficient is as follows:

[0095]

[0096] Among them, η represents the threshold of the cross-correlation coefficient, max(ρ k ) represents the maximum value of the mutual correlation coefficient between the IMF component and the vibration signal;

[0097] A4. Compare the cross-correlation coefficient with the standard deviation of the cross-correlation coefficient. If the cross-correlation coefficient is greater than the threshold of the cross-correlation coefficient, the IMF component corresponding to the cross-correlation coefficient is taken as the effective IMF component.

[0098] A5. Reconstruct the effective IMF components based on singular spectrum analysis to obtain the noise-reduced vibration signal.

[0099] In this embodiment, variational mode decomposition is a signal decomposition technology, which separates the various frequency components of the signal by decomposing the signal into multiple intrinsic mode functions (IMFs). Each IMF component contains a relatively independent oscillation mode. VMD can process nonlinear and non-stationary signals and is suitable for complex time-varying signals such as vibration signals. In signal decomposition, IMF components refer to signal components decomposed by a specific algorithm (such as VMD). Each IMF component reflects different vibration modes or frequency components in the signal. Generally, IMF has a zero mean and has a single extreme point in its rising and falling cycles. The mutual correlation coefficient is an indicator to measure the similarity or correlation between two signals (such as IMF components and vibration signals). It can describe the similarity of the two signals under a certain time delay. The value ranges from -1 (complete negative correlation) to +1 (complete positive correlation), and 0 indicates no correlation. Singular spectrum analysis is a time series analysis method based on matrix decomposition, which is used to extract the main modes in the signal. SSA identifies the main components in the signal and removes noise by embedding and singular value decomposition of the time series. For effective I MF component, SSA can be used to reconstruct the denoised signal.

[0100] In an optional embodiment, the expression of windowing processing is as follows:

[0101]

[0102] Wherein, ω(n) represents the windowing process of the noise reduction vibration signal, L represents the window length of the windowing window, and n represents the total length of the noise reduction vibration signal;

[0103] The calculation formula of short-time vibration energy value is as follows:

[0104]

[0105] Wherein, E(t) represents the short-time vibration energy value of the short-time vibration signal, [t1, t2] represents the time interval of the short-time vibration signal, and x(t) represents the short-time vibration signal.

[0106] In an optional embodiment, the action time of the short-time vibration energy value sequence is extracted based on the double threshold method to obtain the action time sequence, including:

[0107] B1. Preset a low threshold value, and identify whether the short-time vibration signal corresponding to the short-time vibration energy value has entered the starting state based on the low threshold value. If the short-time vibration energy value is greater than the low threshold value, it indicates that it is in the starting state;

[0108] B2. Preset a high threshold value, and identify whether the short-time vibration signal corresponding to the short-time vibration energy value enters the termination state based on the high threshold value. If the short-time vibration energy value is greater than the high threshold value, it indicates that it is in the termination state;

[0109] B3. When the short-time vibration energy value is greater than the lower threshold, the time corresponding to the short-time vibration energy value at this time is taken as the initial point;

[0110] B4. Traverse the short-time vibration energy value sequence after the initial point. When the short-time vibration energy value is greater than the upper threshold, the time corresponding to the short-time vibration energy value at this time is taken as the end point, and the action time is obtained based on the initial point and the end point;

[0111] B5. Repeat the above operations until the short-time vibration energy value sequence is traversed to obtain the action time sequence.

[0112] In an optional embodiment, performing life prediction on the action time series based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply includes:

[0113] C1. Define the action time degradation characteristic quantity of the target switching power supply, where the expression of the action time degradation characteristic quantity is as follows:

[0114] X(t)=a+μt+σ B B(t);

[0115] Among them, X(t) represents the degradation characteristic of action time, a represents the initial value of action time, μ represents the drift coefficient, σ B represents the diffusion coefficient, B(t) represents the standard Wiener process, and t represents the action time series;

[0116] C2. Define the relative increment of the action time degradation characteristic quantity, where the relative increment expression is as follows:

[0117] ΔX(t i )=a+μt i +σ B ΔB(t i );

[0118] Among them, ΔX(t i ) represents the relative increment of the action time degradation characteristic quantity, t i represents the i-th action time in the action time series, ΔB(t i ) represents the relative increment of the standard Wiener process;

[0119] C3, calculating the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity based on maximum likelihood estimation;

[0120] C4. Construct the remaining life probability density function of the target switching power supply, substitute the drift parameter and diffusion coefficient of the relative increment of the action time degradation characteristic quantity into the remaining life probability density function, and take the time corresponding to the maximum value of the remaining life probability density function as the first remaining service life of the target switching power supply, wherein the expression of the remaining life probability density function is as follows:

[0121]

[0122] Among them, f(t|μ,σ B ,h) represents the remaining life probability density function, and h represents the failure threshold.

[0123] It should be noted that the degradation feature of action time is a feature used to measure the change in action time of the target switching power supply during use. It is usually closely related to the aging or degradation process of the equipment. The degradation feature can reflect the performance degradation of the power supply, such as the increase in action time. The initial value refers to the initial value of the degradation feature of the action time when the equipment just starts to run. It is usually a known reference value, representing the health status or good working condition of the equipment. The drift coefficient refers to the rate of change of the degradation feature over time, that is, how the action time shifts over time. The diffusion coefficient is a parameter that measures the random fluctuations of the degradation process, which represents the randomness or uncertainty of the system and controls the amplitude of the change in the action time. The standard Wiener process Process) is a continuous time random process, which describes the random fluctuations of the system and is often used to model the randomness of the degradation process; relative increment refers to the ratio of the change of the action time degradation characteristic quantity relative to its previous moment within a period of time; maximum likelihood estimation is a commonly used parameter estimation method, which estimates the model parameters by maximizing the likelihood function of the observed data; the remaining life probability density function is a probability distribution function that describes how long the equipment can operate normally after the current moment.

[0124] In an optional embodiment, the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity are calculated based on the maximum likelihood estimation, including:

[0125] D1. Calculate the likelihood function of the relative increment of the action time degradation feature quantity, where the expression of the likelihood function is as follows:

[0126]

[0127] Among them, L(μ,σ B 2 ) represents the likelihood function of the relative increment of the action time degradation feature, I represents the total number of action times in the action time series, g(ΔX(t i ); μ, σ B 2 ) represents the probability density function of the relative increment of the action time degradation feature quantity;

[0128] D2. Solve the likelihood function of the relative increment of the action time degradation feature quantity to obtain the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity, wherein the calculation formulas of the drift parameter and the diffusion coefficient are as follows:

[0129]

[0130] in, represents the estimated value of the drift parameter, represents an estimate of the diffusion coefficient.

[0131] It should be noted that the likelihood function is a core concept in statistical inference, which is used to describe the possibility of model parameters given observed data. The likelihood function reflects the estimation of parameters by calculating the probability of observed data occurring given model parameters.

[0132] In an optional embodiment, constructing a second life prediction model corresponding to the first remaining useful life sequence based on a grey model includes:

[0133] E1. Accumulate the first remaining useful life sequence to obtain a life accumulation sequence, wherein the accumulation formula is as follows:

[0134]

[0135] Among them, x (1) (k) represents the kth lifetime accumulation in the lifetime accumulation sequence, x (0) (k) represents the kth first remaining useful life in the first remaining useful life sequence;

[0136] E2. Establish a grey differential equation for the life accumulation sequence, where the expression of the grey differential equation is as follows:

[0137] x (0) (k)+αz (1) (k) = β;

[0138] in, α represents the development coefficient of the grey differential equation, β represents the grey action of the grey differential equation;

[0139] E3. Transpose the grey differential equation to obtain the grey prediction model, where the expression of the grey prediction model is as follows:

[0140]

[0141] It should be noted that.

[0142] In an optional embodiment, constructing a second life prediction model corresponding to the first remaining useful life sequence based on the grey model further includes:

[0143] E4. Calculate the development coefficient and grey action of grey prediction model based on the least square method;

[0144] E5. Substitute the development coefficient and the grey action into the grey prediction model, and solve the grey prediction model after substitution to obtain the solution of the grey prediction model after substitution, wherein the expression of the solution of the grey prediction model after substitution is as follows:

[0145]

[0146] in, represents the solution of the grey prediction model after substitution;

[0147] E6. Restore the solution of the grey prediction model after substitution, and obtain a second life prediction model based on the restored grey prediction model.

[0148] It should be noted that the life accumulation sequence is obtained by accumulating the various data in the first remaining useful life sequence; the grey system theory is a theoretical method for dealing with incomplete information and uncertainty problems, and the grey differential equation is the core part of the grey system model, which describes the law of change of the state of a system during its evolution; the grey prediction model is a model based on the grey system theory and used to predict the future state of the system; the least squares method is a mathematical optimization method, which is usually used to fit the model on a given data set. By minimizing the sum of squared errors between data points and model prediction values, the least squares method can find the best fitting parameters of the model; the solution of the grey prediction model after substitution refers to the predicted value of the future behavior of the system obtained by substituting the parameters calculated by the least squares method into the grey prediction model; restoration refers to restoring the solution of the grey prediction model from the accumulation sequence to the original life sequence, because the original data is usually accumulated when constructing the grey prediction model, so after the prediction, the result must be restored in order to obtain the actual first remaining useful life prediction.

[0149] In an optional embodiment, the expression of the second life prediction model is as follows:

[0150]

[0151] in, represents the second life prediction model, represents the restored grey prediction model, and x (0) (1) represents the first remaining useful life in the first remaining useful life sequence.

[0152] Embodiment three, as Figure 2 As shown, the present invention proposes a system for predicting the remaining service life of a switching power supply, which is applicable to the method for predicting the remaining service life of a switching power supply, comprising:

[0153] The signal noise reduction unit 1 is used to obtain a vibration signal of the target switching power supply when the switch is opened, and to reduce the noise of the vibration signal based on variational mode decomposition to obtain a noise-reduced vibration signal;

[0154] Energy calculation unit 2, energy calculation unit 2 is used to perform windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and calculate the short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence;

[0155] A first prediction unit 3, the first prediction unit 3 is used to extract the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and perform life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply;

[0156] The model building unit 4 is used to extract the action time of the short-time vibration energy value sequence based on the double threshold method to obtain the action time sequence, and perform life prediction on the action time sequence based on the pre-trained first life prediction model to obtain the first remaining service life of the target switching power supply;

[0157] The second prediction unit 5 is used to perform life prediction on the short-circuit current based on a second life prediction model to obtain a second remaining service life of the target switching power supply.

[0158] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A method for predicting the remaining useful life of a switching power supply, characterized in that: include: Acquire a vibration signal of a target switching power supply when the switch is opened, and perform noise reduction on the vibration signal based on variational mode decomposition to obtain a noise-reduced vibration signal; Performing windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and calculating the short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence; Extracting the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and performing life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply; Acquire a first remaining service life sequence under different short-circuit currents based on the first service life prediction model, and construct a second service life prediction model corresponding to the first remaining service life sequence based on a grey model; The short-circuit current is predicted for life based on the second life prediction model to obtain a second remaining service life of the target switching power supply.

2. A method for predicting the remaining useful life of a switching power supply according to claim 1, characterized in that: Denoising the vibration signal based on variational mode decomposition to obtain a denoised vibration signal, including: Decomposing the vibration signal into a plurality of IMF components based on variational mode decomposition; Calculate the mutual correlation coefficient between the multiple IMF components and the vibration signal, wherein the calculation formula of the mutual correlation coefficient is as follows: Among them, ρ k represents the mutual correlation coefficient between the IMF component and the vibration signal, N represents the number of sampling points, and f n Indicates vibration signal, represents the mean value of the vibration signal, u kn represents the kth IMF component, represents the mean of the kth IMF component; Calculate the threshold of the mutual correlation coefficient, wherein the calculation formula of the threshold of the mutual correlation coefficient is as follows: Among them, η represents the threshold of the cross-correlation coefficient, max(ρ k ) represents the maximum value of the mutual correlation coefficient between the IMF component and the vibration signal; Comparing the mutual correlation coefficient with the standard deviation of the mutual correlation coefficient, if the mutual correlation coefficient is greater than the threshold of the mutual correlation coefficient, taking the IMF component corresponding to the mutual correlation coefficient as the effective IMF component; The effective IMF components are reconstructed based on singular spectrum analysis to obtain a noise-reduced vibration signal.

3. A method for predicting the remaining useful life of a switching power supply according to claim 2, characterized in that: The expression of the windowing process is as follows: Wherein, ω(n) represents the windowing process of the noise reduction vibration signal, L represents the window length of the windowing window, and n represents the total length of the noise reduction vibration signal; The calculation formula of the short-time vibration energy value is as follows: Wherein, E(t) represents the short-time vibration energy value of the short-time vibration signal, [t1, t2] represents the time interval of the short-time vibration signal, and x(t) represents the short-time vibration signal.

4. A method for predicting the remaining useful life of a switching power supply according to claim 3, characterized in that: Extracting the action time of the short-time vibration energy value sequence based on the double threshold method to obtain the action time sequence includes: Presetting a low threshold value, and identifying whether the short-time vibration signal corresponding to the short-time vibration energy value enters a starting state based on the low threshold value, and if the short-time vibration energy value is greater than the low threshold value, it indicates that it is in the starting state; A high threshold is pre-set, and based on the high threshold, it is identified whether the short-time vibration signal corresponding to the short-time vibration energy value enters a termination state, and if the short-time vibration energy value is greater than the high threshold, it indicates that it is in a termination state; When the short-time vibration energy value is greater than the lower threshold, the time corresponding to the short-time vibration energy value at this time is taken as the initial point; The short-time vibration energy value sequence is traversed after the initial point, and when the short-time vibration energy value is greater than the upper threshold, the time corresponding to the short-time vibration energy value at this time is taken as the end point, and the action time is obtained based on the initial point and the end point; The above operation is repeatedly performed until the short-time vibration energy value sequence is traversed to obtain the action time sequence.

5. A method for predicting the remaining useful life of a switching power supply according to claim 4, characterized in that: Performing a life prediction on the action time series based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply includes: The action time degradation characteristic quantity of the target switching power supply is defined, wherein the expression of the action time degradation characteristic quantity is as follows: X(t)=a+μt+σ B B(t); Among them, X(t) represents the degradation characteristic of action time, a represents the initial value of action time, μ represents the drift coefficient, σ B represents the diffusion coefficient, B(t) represents the standard Wiener process, and t represents the action time series; Define the relative increment of the action time degradation characteristic quantity, wherein the relative increment expression is as follows: ΔX(t i )=a+μt i +s B ΔB(t i ); Among them, ΔX(t i ) represents the relative increment of the action time degradation characteristic quantity, t i represents the i-th action time in the action time series, ΔB(t i ) represents the relative increment of the standard Wiener process; Calculating the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity based on maximum likelihood estimation; Construct a remaining life probability density function of the target switching power supply, substitute the drift parameter and diffusion coefficient of the relative increment of the action time degradation characteristic quantity into the remaining life probability density function, and use the time corresponding to the maximum value of the remaining life probability density function as the first remaining service life of the target switching power supply, wherein the expression of the remaining life probability density function is as follows: Among them, f(t|μ,σ B ,h) represents the remaining life probability density function, and h represents the failure threshold.

6. A method for predicting the remaining useful life of a switching power supply according to claim 5, characterized in that: Calculating the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity based on maximum likelihood estimation includes: Calculate the likelihood function of the relative increment of the action time degradation feature quantity, wherein the expression of the likelihood function is as follows: Among them, L(μ,σ B 2 ) represents the likelihood function of the relative increment of the action time degradation feature, I represents the total number of action times in the action time series, g(ΔX(t i ); μ, σ B 2 ) represents the probability density function of the relative increment of the action time degradation feature quantity; The likelihood function of the relative increment of the action time degradation feature quantity is solved to obtain the drift parameter and diffusion coefficient of the relative increment of the action time degradation feature quantity, wherein the calculation formula of the drift parameter and the diffusion coefficient is as follows: in, represents the estimated value of the drift parameter, represents an estimate of the diffusion coefficient.

7. A method for predicting the remaining useful life of a switching power supply according to claim 1, characterized in that: Constructing a second life prediction model corresponding to the first remaining useful life sequence based on a grey model, including: The first remaining useful life sequence is accumulated to obtain a life accumulation sequence, wherein the accumulation formula is as follows: Among them, x (1) (k) represents the kth lifetime accumulation in the lifetime accumulation sequence, x (0) (k) represents the kth first remaining useful life in the first remaining useful life sequence; A grey differential equation is established for the life accumulation sequence, wherein the expression of the grey differential equation is as follows: x (0) (k)+αz (1) (k)=β; in, α represents the development coefficient of the grey differential equation, β represents the grey action of the grey differential equation; The grey differential equation is transposed to obtain a grey prediction model, wherein the grey prediction model is expressed as follows:

8. A method for predicting the remaining useful life of a switching power supply according to claim 7, characterized in that: Constructing a second life prediction model corresponding to the first remaining useful life sequence based on the grey model, further comprising: Calculating the development coefficient and grey action amount of the grey prediction model based on the least square method; Substitute the development coefficient and the gray action into the gray prediction model, and solve the gray prediction model after substitution to obtain the solution of the gray prediction model after substitution, wherein the expression of the solution of the gray prediction model after substitution is as follows: in, represents the solution of the grey prediction model after substitution; The solution of the grey prediction model after substitution is restored, and a second life prediction model is obtained based on the restored grey prediction model.

9. A method for predicting the remaining useful life of a switching power supply according to claim 8, characterized in that: The expression of the second life prediction model is as follows: in, represents the second life prediction model, represents the restored grey prediction model, and x (0) (1) represents the first remaining useful life in the first remaining useful life sequence.

10. A system for predicting the remaining useful life of a switching power supply, which is applicable to a method for predicting the remaining useful life of a switching power supply as claimed in any one of claims 1 to 9, characterized in that: include: A signal noise reduction unit (1), the signal noise reduction unit (1) is used to obtain a vibration signal of a target switching power supply when the switch is open, and to reduce the noise of the vibration signal based on variational mode decomposition to obtain a noise-reduced vibration signal; An energy calculation unit (2), the energy calculation unit (2) being used to perform windowing processing on the noise reduction vibration signal to obtain a short-time vibration signal sequence, and to calculate the short-time vibration energy value of each short-time vibration signal in the short-time vibration signal sequence to obtain a short-time vibration energy value sequence; A first prediction unit (3), the first prediction unit (3) being used for extracting the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and performing life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply; A model building unit (4), the model building unit (4) is used to extract the action time of the short-time vibration energy value sequence based on a double threshold method to obtain an action time sequence, and perform life prediction on the action time sequence based on a pre-trained first life prediction model to obtain a first remaining service life of the target switching power supply; A second prediction unit (5), the second prediction unit (5) is used to perform life prediction on the short-circuit current based on the second life prediction model to obtain a second remaining service life of the target switching power supply.

Images