A rotary machine residual service life prediction method, device and terminal equipment

Abstract

The application is suitable for the technical field of rotating machinery monitoring, and provides a rotating machinery residual service life prediction method, device and terminal equipment. The method comprises the following steps: reconstructing state monitoring original data of a target rotating machinery based on a deep residual shrinkage network to obtain HI data; and after smoothing the HI data, the residual service life of the target rotating machinery is predicted through an improved HoltDEM. The application adopts APCSSA to optimize and determine the observation weighting parameter and the trend smoothing parameter of the HoltDEM, solves the problem that the traditional HoltDEM super parameter is difficult to determine, and makes it difficult to be put into the long-term changing prediction scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rotating machinery monitoring technology, and particularly relates to a rotating machinery remaining useful life prediction method, device and terminal equipment. BACKGROUND

[0002] Transportation, mining, petrochemical industry, energy and power, national defense industry and the like all rely on a variety of mechanical equipment to coordinate operation to carry out engineering operations, among which, rotating machinery is most widely used, accounting for about 80% of the total number of industrial system equipment. Rotating machinery will continuously degrade, that is, rotating machinery has a service life, therefore, predictive maintenance of rotating machinery is needed to timely replace and repair to avoid causing huge economic losses and personnel accidents.

[0003] At present, the RUL (Remaining useful life) prediction method of rotating machinery can be divided into a model-driven method, a data-driven method and a hybrid-driven method composed of the two. Among them, the model-driven method establishes a mathematical or physical model to describe the equipment degradation information according to the failure mechanism of the equipment, but this method needs to deeply understand the internal structure and mechanism characteristics of the equipment, which is difficult, and the hybrid-driven method also has the same problem; the data-driven method acquires a large amount of data reflecting the equipment health information to construct a model by means of sensors to monitor the operation of the equipment, which has the advantages of not needing to consider mechanical structure, working condition and failure mechanism and the like, and greatly improves the accuracy of the remaining life prediction. In the data-driven method, the Holt DEM is used to predict the RUL of the rotating machinery, which has the characteristics of simple structure, less time and memory consumption and lower requirement for hardware, but the Holt DEM has the problem of difficult determination of hyperparameters, which makes it difficult to be put into a long-term changing prediction scene. SUMMARY

[0004] The main purpose of the present application is to provide a rotating machinery remaining useful life prediction method, device and terminal equipment, which solves the problem that the hyperparameters of the traditional Holt DEM are difficult to determine, making it difficult to be put into a long-term changing prediction scene.

[0005] To achieve the above purpose, the first aspect of the embodiment of the present application provides a rotating machinery remaining useful life prediction method, comprising:

[0006] Reconstructing state monitoring original data of a target rotating machinery based on a deep residual shrinkage network to obtain HI data;

[0007] After smoothing the HI data, predicting the remaining useful life of the target rotating machinery by means of an improved Holt DEM;

[0008] The improved Holt DEM comprises: adopting the APCSSA to optimize and determine the observation weighting parameter and the trend smoothing parameter of the Holt DEM.

[0009] Optionally, the deep residual contraction network is used to reconstruct state monitoring original data of the rotating machinery.

[0010] The HI label is added to the state monitoring original data to obtain state monitoring label data.

[0011] The state monitoring label data is input into the DRSN network for feature extraction to obtain the HI data.

[0012] Optionally, after the HI label is added to the state monitoring original data, the following steps are further included.

[0013] The state monitoring original data to which the HI label is added is subjected to batch standardization and batch normalization processing to obtain the state monitoring label data.

[0014] Optionally, the position updating formula of the discoverer in the APCSSA is as follows:

[0015]

[0016]

[0017]

[0018] wherein r1 and r2 are adaptive gradient factors, is a position updating formula, f g is a current optimal fitness value; f w is a current worst fitness value;

[0019] The position updating formula of the follower in the APCSSA is as follows:

[0020]

[0021] wherein X p is an optimal position occupied by the discoverer; X w is a current worst position, A is a row of a multi-dimensional matrix, each element of which is randomly assigned as 1 or -1, and n is the number of sparrow populations.

[0022] Optionally, when the sparrow realizes the danger, the alarm raiser sends an alarm signal, and the position updating formula of the alarm raiser is as follows:

[0023]

[0024] wherein X bis the current global optimal position; β is a step control parameter, which is a random number obeying normal distribution with mean value 0 and variance 1; K is a random number in [-1, 1]; f i is the fitness value of the current sparrow; f g and f w are the current global optimal and worst fitness values respectively; ε is a minimum constant.

[0025] Optionally, the smoothing processing is Gaussian smoothing processing.

[0026] Optionally, before reconstructing the state monitoring original data of the target rotating machine based on the deep residual shrinkage network, the method comprises:

[0027] According to the type of the rotating machine and the characteristics of the rotating machine data set, the rotating machine data set is filtered to obtain feature data as the original data.

[0028] The second aspect of the embodiment of the application provides a rotating machine residual service life prediction device, comprising:

[0029] A HI data construction module is configured to reconstruct the state monitoring original data of the target rotating machine based on the deep residual shrinkage network, and obtain HI data.

[0030] A residual service life prediction module is configured to predict the residual service life of the target rotating machine by using the improved Holt DEM after smoothing the HI data.

[0031] The improved Holt DEM comprises: using the APCSSA to optimize and determine the observation weighting parameter and the trend smoothing parameter of the Holt DEM.

[0032] The third aspect of the embodiment of the application provides a terminal device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method provided in the first aspect when executing the computer program.

[0033] The fourth aspect of the embodiment of the application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on the processor to implement the steps of the method provided in the first aspect.

[0034] The rotating machine residual service life prediction method provided by the embodiment of the application optimizes and determines the observation weighting parameter and the trend smoothing parameter of the Holt DEM by using the APCSSA, and then predicts the residual service life of the target rotating machine by using the improved Holt DEM, so that the problem that the super parameter of the traditional Holt DEM is difficult to determine and makes it difficult to be put into a long-term changing prediction scene is solved. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 An implementation flowchart of the rotating machinery residual service life prediction method provided by the embodiment of the present application is shown in the figure;

[0036] Figure 2 The HI curve constructed by the embodiment of the present application is shown in the figure;

[0037] Figure 3 The specific process of the APCSSA optimization Holt DEM provided by the embodiment of the present application is shown in the figure;

[0038] Figure 4 The effect of the APCSSA used by the embodiment of the present application to optimize and determine the observation weighting parameter and the trend smoothing parameter is shown in the figure;

[0039] Figure 5 The rotating machinery residual service life prediction result of the embodiment of the present application taken bearing data as an example is shown in the figure;

[0040] Figure 6 The rotating machinery residual service life prediction result of the embodiment of the present application taken aero-engine data as an example is shown in the figure;

[0041] Figure 7 The component structure schematic diagram of the rotating machinery residual service life prediction device provided by the embodiment of the present application is shown in the figure.

[0042] The implementation of the object of the present application, the functional features and the advantages will be further explained by combining with the embodiments and referring to the drawings. DETAILED DESCRIPTION

[0043] It should be understood that the specific embodiments described herein merely serve to explain the present application and do not serve to limit the present application.

[0044] It should be noted that, in this document, the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0045] In this document, the suffix such as "module", "component" or "unit" used to represent an element is only for the convenience of the description of the present application, and has no specific meaning. Therefore, "module" and "component" can be used interchangeably.

[0046] As Figure 1As shown, this embodiment of the invention provides a method for predicting the remaining service life of rotating machinery, including but not limited to the following steps:

[0047] S101. Reconstruct the original condition monitoring data of the target rotating machinery based on the deep residual shrinkage network to obtain HI data.

[0048] In practical applications, the condition monitoring data of rotating machinery has different characteristics depending on the type of rotating machinery. Therefore, this embodiment of the invention filters the rotating machinery dataset to obtain feature data as the original data based on the type of rotating machinery and the characteristics of the dataset. For example, for the bearing dataset, the monitoring data only contains vibration signals, and the data is unprocessed and its size cannot be matched with the network. In this case, its size needs to be adjusted, and the vibration signal data is selected as the original data for constructing the HI. For aero-engine data, the monitoring data consists of signals from 21 different sensors. In this case, feature selection is required as the original data for constructing the HI.

[0049] In one embodiment, the implementation steps of the above step S101, which involves reconstructing using a deep residual shrinkage network, include:

[0050] S1011. Add HI tags to the raw status monitoring data to obtain status monitoring tag data;

[0051] S1012. Input the status monitoring tag data into the DRSN network for feature extraction to obtain the HI data.

[0052] In practical applications, to accelerate network convergence, after adding HI tags to the raw state monitoring data in step S1011 above, the following steps are also included:

[0053] The raw status monitoring data after adding HI tags is processed by batch standardization and batch normalization to obtain the status monitoring tag data.

[0054] For example, the raw condition monitoring data can be represented as follows:

[0055]

[0056] Where e represents the number of elements in the first time series, and s represents the total number of time series.

[0057] Let q∈[1,s], then after adding the HI label, the HI label l of the qth time series is as shown in equation (1). Then, the training set and the test set are batch standardized and normalized according to equations (2) and (3) respectively, and the data size is reshaped according to the characteristics of the dataset to speed up the network convergence speed and make it match the input layer network of DRSN.

[0058]

[0059]

[0060]

[0061] Where, x BN σ represents the batch-standardized data; μ represents the mean of the data sample; σ represents the standard deviation of the data sample; x NL x represents the magnitude of the normalized data. max and x min These represent the maximum and minimum values ​​for each sampling period.

[0062] The test set data, after batch standardization and normalization, is input into the deep residual shrinking network. Its output, after Gaussian smoothing, is HI, represented as a time series (m1, m2…m…). s Taking rolling bearings and aero engines as examples, the HI curve constructed in this invention is as follows: Figure 2 As shown, there are (a) HI curves for rolling bearings and (b) HI curves for aero-engines. In (a), the horizontal axis represents time in 10 seconds, and the vertical axis represents HI data. In (b), the horizontal axis represents time in revolutions, and the vertical axis represents HI data. In (a) and (b), the dots represent data before smoothing, and the lines represent data after smoothing.

[0063] It should be noted that the deep residual shrinking network needs to be pre-trained. The pre-training process is exemplified as follows: First, training set data and test set data are obtained. The training set data is also processed by adding HI tags and batch normalization, as described above. Then, the initial deep residual shrinking network is trained using the processed training set data and test set data to obtain the deep residual shrinking network in this embodiment of the invention.

[0064] The initial deep residual shrinking network consists of one convolutional layer, four residual blocks, one batch normalization (BN) layer, one ReLU activation layer, one global average pooling layer, and one fully connected layer. The convolutional kernel size is 3x3, the stride is 2, and the number of convolutional kernels is 64. For example, the initial deep residual shrinking network uses the Adam optimizer, with a minimum batch size of 64, a training iteration count of 100, and the loss function chosen as the mean squared error as shown in equation (4):

[0065]

[0066] in, Let y be the predicted value of the q-th sample. q Let q be the true value of the q-th sample, and s be the total number of samples.

[0067] For example, the parameter settings for the initial deep residual shrinkage network are shown in Table 1 below.

[0068]

[0069] Table 1

[0070] S102. After smoothing the HI data, the remaining service life of the target rotating machinery is predicted using the improved Holt DEM.

[0071] In step S102 above, the improved Holt DEM predicts the remaining service life of the target rotating machinery. For example, taking a bearing as the target rotating machinery, its remaining service life at the initial prediction time is 18010s, and when the predicted value reaches the threshold 1, the remaining service life at that time is 23790s. Therefore, the remaining service life of the bearing is 23790-18010=5780s.

[0072] In step S102 above, the improved Holt DEM solves the problem of difficult determination of hyperparameters by using APCSSA to optimize the values ​​of observation weighting parameters and trend smoothing parameters.

[0073] In practical applications, the degradation trend of rotating machinery exhibits a long-term trend but lacks seasonal effects. Therefore, HoltDEM is suitable for predicting the remaining service life of rotating machinery. In HoltDEM, two smoothing parameters determine the accuracy of the prediction results: the observation weighting parameter and the trend smoothing parameter. This embodiment of the invention exemplifies the prediction of trending time series data using the following formulas (5)-(7):

[0074] S t =αX t +(1-α)(S t-1 +T t-1 (5)

[0075] T t =β(S t -S t-1 )+(1-β)T t-1 (6)

[0076]

[0077] Among them, S t Let X be the smoothing term at time t, and T be the trend term. The predicted value is α, which is a parameter weighted by past observations to estimate the level demand of the sequence, and β is a parameter for trend smoothing to estimate the growth trend of the sequence, and (0 < α < 1, 0 < β < 1).

[0078] In this embodiment of the invention, the observation weighting parameters and trend smoothing parameters mentioned above are optimized and set using APCSSA.

[0079] It should be noted that APCSSA is an improved algorithm based on SSA. It optimizes model parameters by simulating the foraging and predator-avoiding behaviors of sparrows. In traditional SSA, the top 20% of sparrows with the best fitness values ​​are selected as the discoverers with the widest foraging range, and their position update rules are as follows:

[0080]

[0081] Where t is the current iteration number and T is the total number of iterations. Let α represent the position information of the i-th sparrow in the d-th dimension at the t-th iteration, where α∈(0,1] is a random number, R2∈[0,1] represents the warning value, ST∈[0.5,1] represents the safety value, Q is a random number that follows a standard normal distribution, and L is a one-row multidimensional matrix.

[0082] In practical applications, the discoverer, as the pioneer and leader of the sparrow population, has the greatest impact on the overall algorithm's performance due to its optimization results. In the traditional SSA discoverer position update formula, when R2 < ST, the sparrow is in a safe zone, and its next position is the product of its current position and a random number. Similarly, when R2 > ST, the sparrow is in a dangerous position, and its next position is the sum of its current position and a random number. The uncertainty of random values ​​in the original formula is too dominant, and it doesn't reflect the close relationship between position updates and fitness values. This limits the sparrow's exploration space and may lead it into local optima. Therefore, this embodiment of the invention introduces adaptive variation factors r1 and r2 to improve the original formula. The position update formula for the discoverer provided in this embodiment of the invention is:

[0083]

[0084]

[0085]

[0086] Among them, f g f is the current optimal fitness value; w This is the worst fitness value at present.

[0087] In the improved formula, when R2 < ST, r1 is smaller in the early stage, enhancing the individual sparrow's search ability; r1 is larger in the later stage, expanding the search range and allowing the sparrow to escape local optima, enabling the discoverer to fully explore within a safe area. When R2 > ST, the discoverer is in a dangerous position; r2 is larger in the early stage, allowing the sparrow to escape danger more quickly and return to the group's safe area; r2 is smaller in the later stage, allowing the sparrow to fully explore within a safe area and improving the overall optimization speed.

[0088] Apart from the discoverer, all the remaining sparrows are considered followers. The position update formula for the followers provided in this embodiment of the invention is as follows:

[0089]

[0090] Among them, X p The best position for the discoverer; X w Let A be the worst position, where A is a single-row multidimensional matrix, and each element is randomly assigned a value of 1 or -1. n is the population size of sparrows.

[0091] In one embodiment, when a sparrow senses danger, the early warning agent issues a warning signal, and the position of the early warning agent is updated as shown in equation (13):

[0092]

[0093] Where X b The current global optimal position; β is the step size control parameter, which is a random number following a normal distribution with a mean of 0 and a variance of 1; K∈[-1,1] is a random number; f i f is the current fitness value of the sparrow. g and f w These are the current best and worst fitness values ​​globally, respectively; ε is a minimum constant to avoid zero in the denominator.

[0094] Therefore, this embodiment of the invention optimizes the observation weighting parameters and trend smoothing parameters of HoltDEM using APCSSA. Specifically, it uses the aforementioned location update formulas for discoverers and followers to optimize the observation weighting parameters and trend smoothing parameters of HoltDEM. Based on this, the specific process of optimizing HoltDEM using APCSSA is as follows: Figure 3 As shown, the specific steps are as follows:

[0095] S1. Initialize the parameters of APCSSA and HoltDEM, including the sparrow population size, maximum number of iterations, optimization dimension, and HoltDEM parameters to be optimized, α and β.

[0096] S2. Use the root mean square error of the predicted and actual values ​​of HoltDEM as the fitness function and calculate the initial value.

[0097] S3. Calculate the warning value, and update the location of the discoverer according to formula (11) based on the size of the warning value.

[0098] S4. Update the position of the followers according to equation (12).

[0099] S5. Update the position of some vigilant units according to formula (13).

[0100] S6. Calculate the fitness value of the sparrow's new position, compare the updated fitness value with the original optimal value, and update the global optimal information.

[0101] S7. Repeat steps (2) to (6) until the number of iterations reaches the termination condition, and output the optimal parameters of Holt DEM.

[0102] To verify the improvement effect of using APCSSA to optimize the observation weighting parameters and trend smoothing parameters provided in the embodiments of the present invention, the benchmark test function Schwefel was selected, and APCSSA was compared with other algorithms, including GWO, PSO, GSA, SSA, CSSA, and FASSA, through simulation experiments.

[0103] The benchmark function Schwefel is shown in equation (14) below:

[0104]

[0105] It should be noted that the optimal value, worst value, average value, standard deviation, and average time were used as the evaluation criteria for algorithm performance. All calculation processes were repeated 30 times, and the integrated development environment for the experiment was Matlab (R2020a). The number of iterations for each algorithm was set to 1000, the initial population size to 100, and the spatial dimension to 30. Based on this, the effectiveness of using APCSSA to optimize the observation weighting parameters and trend smoothing parameters in this embodiment of the invention is demonstrated through… Figure 4 The following table shows a comparison of the (a) 3D plot of the Schwefel function and (b) convergence curves. The (a) 3D plot of the Schwefel function includes the X1 axis, X2 axis, and F2 axis. The horizontal axis of the (b) convergence curve comparison plot represents the number of iterations, and the vertical axis represents the suitability. The optimization results of the baseline function are shown in Table 2 below.

[0106]

[0107]

[0108] Table 2

[0109] like Figure 5 and Figure 6As shown, this embodiment of the invention also uses bearing and aero-engine data as examples to illustrate the prediction results obtained by using the above steps S101 and S102 to predict the remaining service life of the target rotating machinery. The specific values ​​and index comparisons are shown in Tables 3 and 4, respectively. Figure 5 The horizontal axis represents time (in 10 seconds), and the vertical axis represents HI data. Figure 6 For the prediction results of aero-engines, the horizontal axis represents time in revolutions, and the vertical axis represents HI data. For example, Figure 5 In the case where the bearing starts to be predicted at 18010s, and the predicted value reaches the threshold 1 at 23790s, then RUL is 23790-18010=5780s.

[0110] like Figure 7 As shown, this embodiment of the invention also provides a device for predicting the remaining service life of rotating machinery, comprising:

[0111] HI data construction module 71 is used to reconstruct the original condition monitoring data of the target rotating machinery based on the deep residual shrinkage network to obtain HI data;

[0112] The remaining service life prediction module 72 is used to predict the remaining service life of the target rotating machinery by smoothing the HI data and using an improved Holt DEM.

[0113] Among them, improving Holt DEM includes: using APCSSA to optimize the observation weighting parameters and trend smoothing parameters of Holt DEM.

[0114] This invention also provides a terminal device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements each step of the method for predicting the remaining service life of rotating machinery as described in the above embodiments.

[0115] This invention also provides a storage medium, which is a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the various steps in the method for predicting the remaining service life of rotating machinery as described in the above embodiments.

[0116] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the foregoing embodiments have described the present invention in detail, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting the remaining service life of rotating machinery, characterized in that, include: The original condition monitoring data of the target rotating machinery is reconstructed based on a deep residual shrinkage network to obtain HI data. After smoothing the HI data, the remaining service life of the target rotating machinery is predicted using an improved Holt DEM. Among them, improving Holt DEM includes: using APCSSA to optimize and set values ​​for the observation weighting parameters and trend smoothing parameters of Holt DEM; The location update formula for the discoverer in APCSSA is as follows: ； (9) ； (10) ； (11) Where r1 and r2 are adaptive variation factors, For the position update formula, This is the current optimal fitness value; t is the current worst fitness value; t is the current iteration number; T is the total number of iterations; R2∈[0,1] represents the warning value, ST∈[0.5,1] represents the safe value, Q is a random number following a standard normal distribution; i is a natural number representing the sparrow position order; The formula for updating the position of followers in APCSSA is: ； (12) in, The best location for the discoverer; This is the worst position at present. Given a single-row, multi-dimensional matrix, where each element is randomly assigned the value 1 or -1. This refers to the sparrow population size. When a sparrow senses danger, the early warning bird issues a warning signal. The early warning bird's position update formula is: ； (13) in This is the current globally optimal position; The step size control parameter is a random number that follows a normal distribution with a mean of 0 and a variance of 1. It is a random number; This represents the current fitness value of the sparrow. and These are the current best and worst fitness values ​​globally, respectively. It is the smallest constant; The specific process for APCSSA optimization of Holt DEM is as follows: S1. Initialize the parameters of APCSSA and Holt DEM, including sparrow population size, maximum number of iterations, optimization dimension, and Holt DEM parameters to be optimized. , ; S2. Use the root mean square error between the predicted and actual values ​​of Holt DEM as the fitness function, and calculate the initial value; S3. Calculate the warning value, and update the location of the discoverer according to formula (11) based on the size of the warning value. S4. Update the position of the followers according to equation (12); S5. Update the positions of some of the vigilant personnel according to formula (13); S6. Calculate the fitness value of the sparrow's new position, compare the updated fitness value with the original optimal value, and update the global optimal information. S7. Repeat steps S2 to S6 until the number of iterations reaches the termination condition, and output the optimal parameters of Holt DEM.

2. The method for predicting the remaining service life of rotating machinery as described in claim 1, characterized in that, The raw condition monitoring data of the target rotating machinery reconstructed based on the deep residual shrinkage network includes: Add HI tags to the raw status monitoring data to obtain status monitoring tag data; The status monitoring tag data is input into the DRSN network for feature extraction to obtain the HI data.

3. The method for predicting the remaining service life of rotating machinery as described in claim 2, characterized in that, After adding HI tags to the raw status monitoring data, the following is included: The raw status monitoring data after adding HI tags is processed by batch standardization and batch normalization to obtain the status monitoring tag data.

4. The method for predicting the remaining service life of rotating machinery as described in claim 1, characterized in that, The smoothing process is Gaussian smoothing.

5. The method for predicting the remaining service life of rotating machinery as described in claim 1, characterized in that, Before reconstructing the raw condition monitoring data of the target rotating machinery based on a deep residual shrinkage network, the following are included: Based on the types of rotating machinery and the characteristics of the rotating machinery dataset, the rotating machinery dataset is filtered to obtain feature data as the raw data.

6. A device for predicting the remaining service life of rotating machinery, characterized in that, include: The HI data construction module is used to reconstruct the raw condition monitoring data of the target rotating machinery based on the deep residual shrinkage network to obtain HI data; The remaining service life prediction module is used to predict the remaining service life of the target rotating machinery by smoothing the HI data and using an improved Holt DEM. Among them, improving Holt DEM includes: using APCSSA to optimize and set values ​​for the observation weighting parameters and trend smoothing parameters of Holt DEM; The location update formula for the discoverer in APCSSA is as follows: ； (9) ； (10) ； (11) Where r1 and r2 are adaptive variation factors, For the position update formula, This is the current optimal fitness value; t is the current worst fitness value; t is the current iteration number; T is the total number of iterations; R2∈[0,1] represents the warning value, ST∈[0.5,1] represents the safe value, Q is a random number following a standard normal distribution; i is a natural number representing the sparrow position order; The formula for updating the position of followers in APCSSA is: ； (12) in, The best location for the discoverer; This is the worst position at present. Given a single-row, multi-dimensional matrix, where each element is randomly assigned the value 1 or -1. This refers to the sparrow population size. When a sparrow senses danger, the early warning bird issues a warning signal. The early warning bird's position update formula is: ； (13) in This is the current globally optimal position; The step size control parameter is a random number that follows a normal distribution with a mean of 0 and a variance of 1. It is a random number; This represents the current fitness value of the sparrow. and These are the current best and worst fitness values ​​globally, respectively. It is the smallest constant; The specific process for APCSSA optimization of Holt DEM is as follows: S1. Initialize the parameters of APCSSA and Holt DEM, including sparrow population size, maximum number of iterations, optimization dimension, and Holt DEM parameters to be optimized. , ; S2. Use the root mean square error between the predicted and actual values ​​of Holt DEM as the fitness function, and calculate the initial value; S3. Calculate the warning value, and update the location of the discoverer according to formula (11) based on the size of the warning value. S4. Update the position of the followers according to equation (12); S5. Update the positions of some of the vigilant personnel according to formula (13); S6. Calculate the fitness value of the sparrow's new position, compare the updated fitness value with the original optimal value, and update the global optimal information. S7. Repeat steps S2 to S6 until the number of iterations reaches the termination condition, and output the optimal parameters of Holt DEM.

7. A terminal device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements each step of the method for predicting the remaining useful life of rotating machinery as described in any one of claims 1 to 5.

8. A storage medium, said storage medium being a computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the method for predicting the remaining service life of rotating machinery as described in any one of claims 1 to 5.

Images