A mechanical equipment residual life prediction method, device, equipment and medium

Abstract

The application discloses a mechanical equipment residual life prediction method, device, equipment and medium, and relates to the field of residual life prediction, and the method comprises the following steps: acquiring equipment data of a target mechanical equipment at a current time; a residual life prediction model is constructed and trained; the residual life prediction model comprises a smoothing processing unit, a feature extraction unit and a prediction unit; in the training process of the residual life prediction model, a Bayesian optimization algorithm is used to optimize the hyperparameters of the residual life prediction model; the equipment data of the target mechanical equipment at the current time is input into the trained residual life prediction model to perform residual life prediction, and the residual life of the target mechanical equipment is obtained; and the application improves the accuracy of the residual life prediction result of the mechanical equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of residual life prediction, in particular to a mechanical equipment residual life prediction method, device, equipment and medium. BACKGROUND

[0002] Mechanical equipment is currently widely used in various fields such as industry, aerospace, etc. Most mechanical equipment needs to run for a long time, and internal components will inevitably wear and age, performance will decline, and even failure will occur, which will lead to major accidents and losses. Therefore, timely and accurate prediction of the residual service life of mechanical equipment, and implementation of scientific and reasonable maintenance measures based on the residual life prediction results, have important practical significance for preventing equipment from failing and ensuring that equipment can work safely and stably.

[0003] Deep learning methods have achieved successful applications in many fields in recent years due to their powerful feature learning and representation capabilities. For example: deep belief networks, convolutional neural networks, long short-term memory networks, and Transformer models have been used in device fault diagnosis and state assessment fields. However, the above deep learning models have a large number of hidden layers and complex model structures, and have the disadvantages of poor interpretability and tedious model training process.

[0004] The generalized regression neural network (GRNN) model has a relatively simple structure, and the training of the model does not require iteration. A large number of research results show that compared with the back propagation (BP) neural network and the support vector machine, the GRNN model has faster training speed and better generalization ability, so the GRNN model has wide application prospects in fault diagnosis and state prediction fields. Considering that there are many hyperparameters in general residual life prediction models based on the GRNN model, manual random setting of the values of the hyperparameters of the model often cannot obtain the optimal result, resulting in low prediction result accuracy. SUMMARY

[0005] The purpose of the present application is to provide a mechanical equipment residual life prediction method, device, equipment and medium, which can improve the accuracy of the mechanical equipment residual life prediction result.

[0006] To achieve the above purpose, the present application provides the following solutions.

[0007] In a first aspect, the present application provides a mechanical equipment residual life prediction method, comprising:

[0008] obtaining equipment data of a target mechanical equipment at a current time;

[0009] construct and train the remaining life prediction model; the remaining life prediction model comprises a smoothing processing unit, a feature extraction unit and a prediction unit; in the training process of the remaining life prediction model, a Bayesian optimization algorithm is used to optimize the hyperparameters of the remaining life prediction model;

[0010] inputting the equipment data of the target mechanical equipment at the current time into the trained remaining life prediction model to perform remaining life prediction, to obtain the remaining life of the target mechanical equipment.

[0011] In a second aspect, the present application provides a mechanical equipment remaining life prediction device, comprising:

[0012] an acquisition module configured to acquire equipment data of a target mechanical equipment at a current time;

[0013] a construction module configured to construct and train a remaining life prediction model; the remaining life prediction model comprises a smoothing processing unit, a feature extraction unit and a prediction unit; in the training process of the remaining life prediction model, a Bayesian optimization algorithm is used to optimize the parameters of the remaining life prediction model;

[0014] a prediction module configured to input the equipment data of the target mechanical equipment at the current time into the trained remaining life prediction model to perform remaining life prediction, to obtain the remaining life of the target mechanical equipment.

[0015] In a third aspect, the present application provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the mechanical equipment remaining life prediction method described above.

[0016] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the mechanical equipment remaining life prediction method described above.

[0017] According to the embodiments provided in the present application, the following technical effects are disclosed:

[0018] The present application provides a mechanical equipment remaining life prediction method, device, equipment and medium, by acquiring equipment data of a target mechanical equipment at a current time, inputting the equipment data of the target mechanical equipment at the current time into a trained remaining life prediction model to perform remaining life prediction, the trained remaining life prediction model is provided with optimal hyperparameters, the optimal hyperparameters are obtained by Bayesian optimization algorithm optimization, compared with the hyperparameters set by artificial, the parameter value obtained by Bayesian optimization algorithm optimization is better, based on this, the remaining life of the mechanical equipment is predicted, and the prediction result is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings described below only constitute some embodiments of the present application. For those skilled in the art, other drawings can also be obtained from these drawings without any creative effort.

[0020] Figure 1 An application environment diagram of a mechanical equipment residual life prediction method in an embodiment of the present application;

[0021] Figure 2 A flowchart of a mechanical equipment residual life prediction method provided in an embodiment of the present application;

[0022] Figure 3 A flowchart of another mechanical equipment residual life prediction method provided in an embodiment of the present application;

[0023] Figure 4 A structure diagram of a GRNN model provided in another embodiment of the present application;

[0024] Figure 5 A functional module diagram of a mechanical equipment residual life prediction device provided in an embodiment of the present application;

[0025] Figure 6 A structure diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort fall within the scope of protection of the present application.

[0027] In order to make the purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0028] The mechanical equipment residual life prediction method provided in the embodiments of the present application can be applied to, for example, Figure 1The application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 to process. The data storage system can be set up separately, or integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the device data of the target mechanical equipment at the current time to the server 104, and the server 104 receives the device data of the target mechanical equipment at the current time. For the device data of the target mechanical equipment at the current time, the server 104 constructs and trains the remaining life prediction model; The remaining life prediction model includes a smoothing processing unit, a feature extraction unit and a prediction unit; In the training process of the remaining life prediction model, the hyperparameters of the remaining life prediction model are optimized by using the Bayesian optimization algorithm; Input the device data of the target mechanical equipment at the current time into the trained remaining life prediction model to predict the remaining life, and obtain the remaining life of the target mechanical equipment. The server 104 can feed back the obtained remaining life of the target mechanical equipment to the terminal 102. In addition, in some embodiments, the mechanical equipment remaining life prediction method can also be realized by the server 104 or the terminal 102 alone, such as the terminal 102 can directly predict the remaining life of the target mechanical equipment at the current time. The device data of the target mechanical equipment at the current time can also be obtained from the data storage system by the server 104, and the device data of the target mechanical equipment at the current time can be processed to predict the remaining life.

[0029] Among them, the terminal 102 can be but not limited to various desktop computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices, Internet of Things devices can be smart speakers, smart televisions, smart air conditioners, smart car devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be realized by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0030] In an exemplary embodiment, as shown in Figure 2 and Figure 3 A mechanical equipment remaining life prediction method is provided, which is executed by a computer device, specifically by a terminal or a server, etc. Computer device alone, or by a terminal and a server together, in the embodiment of the application, taking the server 104 in Figure 1 as an example for illustration, including the following steps 201 to 203. Among them:

[0031] Step 201, obtaining the device data of the target mechanical equipment at the current time.

[0032] In step 202, a remaining useful life prediction model is constructed and trained; the remaining useful life prediction model comprises a smoothing processing unit, a feature extraction unit and a prediction unit; in the training process of the remaining useful life prediction model, a Bayesian optimization algorithm is used to optimize the hyperparameters of the remaining useful life prediction model.

[0033] In step 203, the equipment data of the target mechanical equipment at the current time is input into the trained remaining useful life prediction model for remaining useful life prediction, and the remaining useful life of the target mechanical equipment is obtained.

[0034] The method of steps 201-203 can improve the accuracy of the remaining useful life prediction result of the mechanical equipment.

[0035] In an exemplary embodiment, step 203 specifically comprises steps 301-303:

[0036] In step 301, the equipment data of the target mechanical equipment at the current time is input into the smoothing processing unit for smoothing filtering processing, and smoothed equipment data is obtained.

[0037] Specifically, the smoothing processing unit uses a mean filtering method to perform smoothing filtering processing on the equipment data of the target mechanical equipment at the current time.

[0038] In industrial automation and control systems, the mean filtering method is widely used to remove noise from data collected by sensors, improving the stability and reliability of the system, so the mean filtering method is used in this embodiment to remove noise from data collected by sensors. Let the collected raw data be Then, the arithmetic mean of the data collected by each sensor is calculated within a window width parameter The calculation formula of the mean filtering method is:

[0039] (1);

[0040] Wherein, is the smoothed equipment data, , m is the data dimension, , is the running time of the mechanical equipment, is the window width parameter, and c is a constant.

[0041] In step 302, the smoothed equipment data is input into the feature extraction unit for feature extraction, and the degradation features of the target mechanical equipment are obtained. Specifically, the feature extraction unit uses a principal component analysis method to perform feature extraction on the smoothed equipment data, and obtains the degradation features of the target mechanical equipment.

[0042] Although the data quality after the smoothing filtering processing is improved compared with the original data, due to the large number of sensors, there are some useless data or data redundancy, and with the increase of data dimension, the number of samples required by algorithm learning increases exponentially, and directly using the device data after smoothing filtering processing is not conducive to the training and use of the remaining life prediction model. Therefore, the embodiment uses the Kernel Principal Component Analysis (KPCA) method to extract features from the device data after the smoothing filtering processing.

[0043] KPCA is a nonlinear data processing method, and its core idea is to project the data in the original space to a high-dimensional feature space through a nonlinear mapping, and then perform data processing based on principal component analysis in the high-dimensional feature space. The algorithm steps for feature extraction based on KPCA are as follows:

[0044] 1. Select the Gaussian radial basis function as the kernel function:

[0045] (2);

[0046] wherein, is the kernel parameter.

[0047] 2. Calculate the kernel matrix , The element in the pth row and qth column of satisfies: , , wherein, is the vector in the pth column of , and is the device data after smoothing filtering processing.

[0048] 3. Calculate the centralized kernel matrix , and the calculation formula is:

[0049] (3);

[0050] wherein, is an n-dimensional matrix with all elements being 1.

[0051] 4. Perform eigenvalue decomposition on the centralized kernel matrix , After sorting the eigenvalue size, the eigenvalue is , and the corresponding eigenvector is .

[0052] 5. Take the first principal component as the degradation feature of the mechanical equipment, wherein is the transpose of the vector.

[0053] Step 303: Input the degradation features into the prediction unit to predict the remaining lifespan and obtain the remaining lifespan of the target mechanical equipment.

[0054] Specifically, the prediction unit uses the GRNN model to predict the remaining life of the target mechanical equipment, thereby obtaining the remaining life of the target mechanical equipment.

[0055] The result obtained in step 302 The actual lifespan of the mechanical equipment is used as the input to the prediction unit to train the GRNN model.

[0056] The GRNN model is a variation of the radial basis function neural network. The theoretical basis of the GRNN model is nonlinear kernel regression analysis, using non-independent variables. Relative to independent variables Regression analysis is actually calculating the value with the highest probability. Let random variable and The joint probability density function is The actual observed value is known. That is, the raw data collected. This refers to all the data collected by the sensors at a certain moment when the mechanical equipment is running, including the low-pressure turbine outlet temperature of the engine, the fan inlet pressure, the uncorrected core engine speed, the corrected core engine speed, the bypass ratio, the total pressure of the outer bypass duct, the engine pressure ratio, the high-pressure compressor outlet static pressure, and the corrected fan speed.

[0057] The training samples consist of degradation features and the corresponding actual values ​​of remaining lifetime. Compared to The regression, i.e., the conditional mean, is:

[0058] (4);

[0059] For an unknown probability density function , Degenerative characteristics The actual value of the corresponding remaining lifespan.

[0060] (5);

[0061] in: and Let be the degradation feature corresponding to the i-th training sample and the actual value of the remaining lifetime corresponding to the i-th training sample. As a smoothing factor, The number of training samples is used. Replace the formula (4) Calculation yields:

[0062] (6);

[0063] remember:

[0064] (7);

[0065] in, It measures the input vector With degenerative characteristics The similarity between the input space and the feature space is used to complete the nonlinear mapping from the input space to the high-dimensional feature space.

[0066] (8);

[0067] (9);

[0068] Among them, the summation layer has a total of 1 neuron, including indivual Neuron and one Neuron, To find the first layer One neuron, Degenerative characteristics The actual value of the corresponding remaining lifetime, the sum of the weighted outputs of all pattern layer neurons. .

[0069] (10);

[0070] in, This is the sum of the outputs of all neurons in the pattern layer.

[0071] Then formula (6) can be simplified to:

[0072] (11);

[0073] Formulas (4) to (11) can be described by the GRNN four-layer network structure model. The GRNN network structure is as follows: Figure 4 As shown.

[0074] Raw data collected After smoothing filtering and KPCA feature extraction in steps 301 and 302, the degradation features of the mechanical equipment are obtained, which is the input vector. , the output result is obtained after the degenerative feature input GRNN network sequentially passes through the input layer, the mode layer, the summation layer and the output layer, wherein the number of neurons of the input layer is equal to the vector dimension in the learning sample; the number of neurons of the mode layer is equal to the number of learning samples.

[0075] The input layer receives the input vector and directly transmits it to the mode layer. There is no actual calculation operation in this layer, only data transmission.

[0076] The mode layer has neurons, each of which corresponds to a training sample. For the th neuron, the input is , and the output is . The mode layer calculates the similarity of the input vector and each training sample vector through a Gaussian function, completing the nonlinear mapping from the input space to the high-dimensional feature space.

[0077] Let the connection coefficient of the mode layer and the neuron of the summation layer be 1, and the connection coefficient of the mode layer and the remaining neurons of the summation layer be , and the final output is obtained through formula (8) to formula (11).

[0078] In an exemplary embodiment, the remaining life prediction model includes a plurality of hyperparameters, the hyperparameters including a window width parameter , a kernel parameter and a smoothing factor , the window width parameter , the kernel parameter and the smoothing factor constitute a hyperparameter combination, and the hyperparameters of the remaining life prediction model are optimized in step 202 using a Bayesian optimization (BO) algorithm, specifically including steps 401-407:

[0079] Step 401, define a hyperparameter combination space, and randomly select multiple groups of hyperparameter combinations from the hyperparameter combination space as initial hyperparameter combinations; the hyperparameter combination space includes multiple groups of hyperparameter combinations.

[0080] Step 402, construct a target function with the minimum average absolute error between the predicted remaining life and the actual remaining life output by the remaining life prediction model in the training process as the target, and calculate the target function value of each initial hyperparameter combination through the target function.

[0081] Specifically, the data sampled by the sensor is divided into a training set , a validation set and a test set finding a set of hyperparameter combinations in the hyperparameter combination space , so that the average absolute error of the remaining life prediction results of the validation set is minimized:

[0082] (12);

[0083] wherein, is the objective function, is the average absolute error of the remaining life prediction results of the validation set.

[0084] Step 403, constructing an initial data set, the initial data set comprising a plurality of initial data points, the initial data points comprising initial hyperparameter combinations and corresponding objective function values.

[0085] Specifically, randomly sampling initial hyperparameter combinations: randomly selecting init hyperparameter combinations in the hyperparameter combination space , running the objective function for each initial hyperparameter combination to obtain the corresponding objective function value, and constructing the initial data set ; wherein, is the i th initial hyperparameter combination, is the objective function value of the i th initial hyperparameter combination.

[0086] Step 404, constructing a Gaussian proxy model based on the initial data set; the Gaussian proxy model is used to estimate the distribution of the objective function in the entire hyperparameter space according to the initial data points.

[0087] For init initial hyperparameter combinations , the corresponding objective function values are , and it is assumed that obeys a Gaussian distribution, that is: .

[0088] wherein, is the mean vector of the Gaussian process prediction, is a calculation function of the mean, and is initially a zero vector.

[0089] Covariance matrix: (13);

[0090] The covariance function is used to measure the similarity between the th initial hyperparameter combination and the th hyperparameter combination .

[0091] Step 405, find the super parameter combination that makes the expected improvement function take the maximum value as the new super parameter combination in the whole super parameter combination space through the gradient ascent algorithm, and evaluate the new super parameter combination using the objective function to obtain the objective function value of the new super parameter combination; the expected improvement function is composed of the minimum value in the objective function value in the initial data set and the variance of the Gaussian surrogate model.

[0092] Specifically, the next evaluation point is selected based on the expected improvement function:

[0093] For a Gaussian process, a new super parameter combination Its predicted posterior distribution Also conforms to the Gaussian distribution. For a certain new super parameter combination , the corresponding objective function value is predicted as .

[0094] According to the conditional probability property of the Gaussian process, the mean value and the covariance matrix of the posterior distribution are respectively:

[0095] (14);

[0096] (15);

[0097] The calculation formula of the variance of the posterior distribution is:

[0098] (16);

[0099] Record the current optimal value of the objective function in the init initial super parameter combinations , ,…, :

[0100] (17);

[0101] Use the expected improvement function (Expected Improvement, EI) as the acquisition function. For a new super parameter combination , the expected improvement function is:

[0102] (18);

[0103] (19);

[0104] Where: is the expectation; and Cumulative distribution function (CDF) and probability density function (PDF) of standard normal distribution, respectively.

[0105] Based on the above principle, the point of maximum value of the expected promotion function is found by gradient ascent algorithm As the next evaluation point, that is, the new hyperparameter combination.

[0106] In the selected new hyperparameter combination The objective function is evaluated at .

[0107] Step 406, add the new hyperparameter combination and the objective function value of the new hyperparameter combination to the initial data set as a new data point, and iteratively update the mean and variance of the Gaussian proxy model until the maximum number of iterations is reached.

[0108] Specifically, the new hyperparameter combination is added to the initial data set: ; Wherein, The updated data set.

[0109] Step 407, select the hyperparameter combination with the minimum objective function value from all evaluated hyperparameter combinations as the optimal hyperparameter combination.

[0110] Finally, the remaining life prediction performance of the constructed remaining life prediction model is verified by the test set .

[0111] Based on the same inventive concept, the embodiments of the present application also provide a mechanical equipment remaining life prediction device for implementing the mechanical equipment remaining life prediction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more mechanical equipment remaining life prediction device embodiments provided below can refer to the limitations of the mechanical equipment remaining life prediction method in the above, which will not be repeated here.

[0112] In one exemplary embodiment, as Figure 5 shown, a mechanical equipment remaining life prediction device is provided, comprising:

[0113] The acquisition module 51 is configured to acquire equipment data of the target mechanical equipment at the current time.

[0114] The construction module 52 is configured to construct and train a remaining life prediction model; the remaining life prediction model comprises a smoothing processing unit, a feature extraction unit and a prediction unit; in the training process of the remaining life prediction model, a Bayesian optimization algorithm is used to optimize the parameters of the remaining life prediction model.

[0115] The prediction module 53 is configured to input the equipment data of the target mechanical equipment at the current time into the trained remaining life prediction model to perform remaining life prediction, and obtain the remaining life of the target mechanical equipment.

[0116] In an exemplary embodiment, a computer device is provided, which can be a server or a terminal, and an internal structure diagram thereof can be as shown in Figure 6 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is configured to store equipment data of a target mechanical equipment at a current time. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement the mechanical equipment remaining life prediction method.

[0117] Those skilled in the art can understand that Figure 6 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0118] In an exemplary embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0119] In an exemplary embodiment, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the steps in the above method embodiments.

[0120] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0121] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to a memory, a database or other medium used in the embodiments provided in the present application can include at least one of a non-volatile and a volatile memory. The non-volatile memory can include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical storage, a high-density embedded non-volatile memory, a resistive random access memory (ReRAM), a magnetoresistive random access memory (MRAM), a ferroelectric random access memory (FRAM), a phase change memory (PCM), a graphene memory, etc. The volatile memory can include a random access memory (RAM) or an external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), etc.

[0122] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0123] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.

[0124] The principles and implementation modes of the present application are described by applying specific examples in the present application. The above-mentioned embodiments are only used to help understand the method and its core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range can be changed. In conclusion, the content of the present application should not be understood as a limitation.

Claims

1. A method for predicting the remaining life of a mechanical device, characterized by, The method comprises the following steps: acquiring equipment data of a target mechanical device at a current time; the equipment data comprises engine low-pressure turbine outlet temperature, fan inlet pressure, uncorrected core engine speed, core engine corrected speed, bypass ratio, outer bypass total pressure, engine pressure ratio, high-pressure compressor outlet static pressure, and fan corrected speed; constructing and training a remaining useful life prediction model; the remaining useful life prediction model comprises a smoothing processing unit, a feature extraction unit, and a prediction unit; a Bayesian optimization algorithm is used to optimize hyperparameters of the remaining useful life prediction model in the training process; inputting the equipment data of the target mechanical device at the current time into the trained remaining useful life prediction model to predict the remaining useful life of the target mechanical device; specifically, the smoothing processing unit uses a mean filtering method to perform smoothing filtering processing on the equipment data of the target mechanical device at the current time; the feature extraction unit uses a principal component analysis method to extract features from the smoothed equipment data to obtain degradation features of the target mechanical device; and the prediction unit uses a GRNN model to predict the remaining useful life of the target mechanical device to obtain the remaining useful life of the target mechanical device; the hyperparameters of the remaining useful life prediction model comprise a window width parameter of the mean filtering method, a kernel parameter of the principal component analysis method, and a smoothing factor of the GRNN model; the calculation formula of the mean filtering method is as follows: wherein, is the smoothed equipment data, h = 1, …, m, m is the data dimension, j = 1, …, r, r is the time of mechanical equipment operation, γ is the window width parameter, and c is a constant. the prediction unit uses the GRNN model to predict the remaining useful life of the target mechanical device, specifically including: after the degradation features are input into the GRNN network, the degradation features sequentially pass through an input layer, a pattern layer, a summation layer, and an output layer to obtain an output result; the number of neurons in the input layer is equal to the vector dimension in the learning sample; the number of neurons in the pattern layer is equal to the number of learning samples; the input layer receives an input vector α and directly transmits the input vector α to the pattern layer; The mode layer has n neurons, each neuron corresponds to a training sample, and the input of the i-th neuron is a and the output is P i (a); the mode layer calculates the similarity between the input vector and each training sample vector through a Gaussian function, and completes the nonlinear mapping from the input space to the high-dimensional feature space; The summation layer has n+1 neurons, including n neurons and one S D neuron, is the i+1th neuron of the summation layer; Y i is the degradation feature α i corresponding to the actual value of the remaining life; S D is the sum of the outputs of all mode layer neurons, S N is the sum of the weighted outputs of all mode layer neurons; the connection coefficient between the mode layer and the summation layer neuron S D is 1, and the connection coefficient between the mode layer and the remaining neurons of the summation layer is Y i The final output is calculated by the following formula wherein, δ is a smoothing coefficient.

2. The mechanical equipment residual life prediction method according to claim 1, characterized by, inputting the equipment data of the target mechanical device at the current time into the trained remaining useful life prediction model to predict the remaining useful life of the target mechanical device, including: inputting the equipment data of the target mechanical device at the current time into the smoothing processing unit to perform smoothing filtering processing and obtain smoothed equipment data; inputting the smoothed equipment data into the feature extraction unit to extract features and obtain degradation features of the target mechanical device; inputting the degradation features into the prediction unit to predict the remaining useful life of the target mechanical device.

3. The mechanical equipment residual life prediction method according to claim 1, characterized by, The remaining useful life prediction model comprises a plurality of hyperparameters, and the plurality of hyperparameters form a hyperparameter combination; wherein, the Bayesian optimization algorithm is used to optimize the hyperparameters of the remaining useful life prediction model, specifically including: defining a hyperparameter combination space, randomly selecting a plurality of hyperparameter combinations from the hyperparameter combination space as initial hyperparameter combinations; the hyperparameter combination space comprises a plurality of hyperparameter combinations; constructing an objective function with the minimum average absolute error between the predicted remaining useful life and the actual remaining useful life output by the remaining useful life prediction model in the training process as the target, and calculating the objective function value of each initial hyperparameter combination through the objective function; Construct an initial data set, the initial data set comprising a plurality of initial data points, the initial data points comprising initial hyperparameter combinations and corresponding objective function values; Construct a Gaussian surrogate model based on the initial data set; the Gaussian surrogate model is used to estimate the distribution of the objective function over the entire hyperparameter combination space according to the initial data points; Find, by using a gradient ascent algorithm, a hyperparameter combination that makes the expected improvement function take the maximum value in the entire hyperparameter combination space as a new hyperparameter combination, and evaluate the new hyperparameter combination by using the objective function to obtain the objective function value of the new hyperparameter combination; the expected improvement function is composed of the minimum value in the objective function values in the initial data set and the variance of the Gaussian surrogate model; Add the new hyperparameter combination and the objective function value of the new hyperparameter combination as a new data point to the initial data set, and iteratively update the mean and variance of the Gaussian surrogate model until a set maximum number of iterations is reached; Select, from all evaluated hyperparameter combinations, a hyperparameter combination with the minimum objective function value as an optimal hyperparameter combination.

4. A mechanical device remaining life prediction apparatus characterized by comprising: Comprise: An acquisition module is configured to acquire equipment data of a target mechanical device at a current time; the equipment data comprises an engine low-pressure turbine outlet temperature, a fan inlet pressure, an uncorrected core engine speed, a core engine corrected speed, a bypass ratio, an outer bypass total pressure, an engine pressure ratio, a high-pressure compressor outlet static pressure, and a fan corrected speed; A construction module is configured to construct and train a remaining useful life prediction model; the remaining useful life prediction model comprises a smoothing processing unit, a feature extraction unit, and a prediction unit; a Bayesian optimization algorithm is used to optimize parameters of the remaining useful life prediction model in a training process of the remaining useful life prediction model; A prediction module is configured to input the equipment data of the target mechanical device at the current time into the trained remaining useful life prediction model to predict a remaining useful life of the target mechanical device; specifically, the smoothing processing unit performs mean filtering on the equipment data of the target mechanical device at the current time; the feature extraction unit extracts degradation features of the target mechanical device by using principal component analysis on the equipment data filtered by the smoothing processing unit; and the prediction unit predicts the remaining useful life of the target mechanical device by using a GRNN model to obtain the remaining useful life of the target mechanical device; the hyperparameters of the remaining useful life prediction model comprise a window width parameter of the mean filtering method, a kernel parameter of the principal component analysis method, and a smoothing factor of the GRNN model; A calculation formula of the mean filtering method is as follows: wherein, is the smoothed equipment data, h = 1, …, m, m is the data dimension, j = 1, …, r, r is the time of mechanical equipment operation, γ is the window width parameter, and c is a constant. The prediction unit predicts the remaining useful life of the target mechanical device by using the GRNN model; specifically, after the degradation features are input into the GRNN network, the degradation features sequentially pass through an input layer, a pattern layer, a summation layer, and an output layer to obtain an output result; the number of neurons in the input layer is equal to the vector dimension in the learning samples; and the number of neurons in the pattern layer is equal to the number of learning samples. The input layer receives an input vector α and directly transmits the input vector α to the pattern layer. The mode layer has n neurons, each neuron corresponds to a training sample, and the input of the i-th neuron is a and the output is P i (a); the mode layer calculates the similarity between the input vector and each training sample vector through a Gaussian function, and completes the nonlinear mapping from the input space to the high-dimensional feature space; The summation layer has n+1 neurons, including n neurons and one S D neuron, is the i+1th neuron of the summation layer; Y i is the degradation feature α i corresponding to the actual value of the remaining life; S D is the sum of the outputs of all mode layer neurons, S N is the sum of the weighted outputs of all mode layer neurons; the connection coefficient between the mode layer and the summation layer neuron S D is 1, and the connection coefficient between the mode layer and the remaining neurons of the summation layer is Y i The final output is calculated by the following formula wherein, δ is a smoothing coefficient.

5. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the mechanical equipment residual life prediction method of any one of claims 1-3.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the mechanical equipment residual life prediction method of any one of claims 1-3.

Images