Method and equipment for predicting residual life of mechanical equipment

Abstract

The invention provides a mechanical equipment residual life prediction method and equipment. The method comprises the following steps: acquiring actual vibration data of target mechanical equipment in a preset time window; performing working condition prediction on the actual vibration data of the preset time window by adopting a target working condition prediction module in a preset health reference model, and determining a prediction probability that the target mechanical equipment belongs to a plurality of preset working conditions in the preset time window; target prediction subnets of multiple preset working conditions in a preset health reference model are adopted to predict the actual vibration data of a preset time window, and vibration data prediction values of the target mechanical equipment at the current moment for the multiple preset working conditions are determined; determining a health index of the target mechanical equipment at the current moment according to the prediction probabilities of the plurality of preset working conditions, the vibration data prediction values of the plurality of preset working conditions and the actual vibration data at the current moment; and determining the remaining service life of the target mechanical equipment at the current moment according to the health index of the target mechanical equipment at the current moment.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method and apparatus for predicting the remaining life of mechanical equipment. Background Technology

[0002] With the development of industry and intelligent operation and maintenance, accurately predicting the remaining service life of critical equipment (such as bearings and gears in rotating machinery) is of great significance for achieving predictive maintenance and ensuring production safety. However, modern industrial equipment often operates under various conditions, and the dynamic characteristics and degradation rates of different conditions often vary significantly. A single model cannot adapt to multiple operating modes simultaneously, resulting in a serious lack of prediction accuracy under varying operating conditions.

[0003] Currently, existing life prediction methods for multi-condition problems first identify the operating conditions of the equipment and then establish an independent prediction model for each condition. However, these methods typically rely on a "hard division" of operating conditions, assuming that the equipment is in only one specific operating condition at any given time.

[0004] This approach has significant limitations: First, in actual operation, equipment operating conditions may switch gradually, or its operating state itself may be a mixture of multiple modes, and hard partitioning introduces the risk of misjudgment. Second, when equipment performance begins to degrade, its operating characteristics often deviate from purely healthy operating conditions and become ambiguous, making it difficult for hard partitioning-based models to accurately classify them, thus leading to inaccurate subsequent predictions. Furthermore, existing methods typically lack a unified health indicator decoupled from specific operating conditions to quantify the overall degree of equipment degradation, resulting in significant fluctuations and insufficient reliability in the final remaining life prediction results. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the prior art by providing a method and device for predicting the remaining life of mechanical equipment. This method and device can improve the accuracy and robustness of equipment remaining life prediction by acquiring actual vibration data within a preset time window and combining it with a health benchmark model to achieve probability prediction of operating conditions, vibration data prediction, health index calculation, and remaining life calculation.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, embodiments of this application provide a method for predicting the remaining life of mechanical equipment, including: Acquire actual vibration data of the target mechanical equipment within a preset time window; the preset time window includes the current moment and a preset time period prior to the current moment. Using the target working condition prediction module in the preset health benchmark model, the actual vibration data of the preset time window is used to predict the working condition and determine the predicted probability that the target mechanical equipment belongs to multiple preset working conditions in the preset time window. Using the target prediction subnet of the multiple preset working conditions in the preset health benchmark model, the actual vibration data of the preset time window is predicted to determine the predicted value of the vibration data of the target mechanical equipment for the multiple preset working conditions at the current time. Based on the predicted probabilities of the multiple preset working conditions, the predicted vibration data of the multiple preset working conditions, and the actual vibration data at the current moment, the health index of the target mechanical equipment at the current moment is determined. Based on the health indicators of the target mechanical equipment at the current moment, determine the remaining service life of the target mechanical equipment at the current moment.

[0007] In an optional implementation, determining the health index of the target mechanical equipment at the current moment based on the predicted probabilities of the plurality of preset working conditions, the predicted vibration data values ​​of the plurality of preset working conditions, and the actual vibration data at the current moment includes: Based on the predicted vibration data of the multiple preset working conditions and the actual vibration data at the current moment, the prediction error of each preset working condition is determined. Based on the prediction error of each preset working condition, the mean parameter and standard deviation parameter of each preset working condition, the standardized score corresponding to each preset working condition is determined. Based on the standardized score corresponding to each preset working condition and the predicted probability of each preset working condition, the health index of the target mechanical equipment at the current moment is determined.

[0008] In an optional implementation, determining the remaining service life of the target mechanical equipment at the current moment based on its health indicators includes: Based on the health indicators of the target mechanical equipment at the current moment and the health indicators at historical moments, determine the performance degradation rate of the target mechanical equipment at the current moment; The remaining service life of the target mechanical equipment at the current moment is determined based on the performance degradation rate, the health index of the target mechanical equipment at the current moment, and the preset health index failure threshold.

[0009] In an optional implementation, the method further includes: If the health indicators of the target mechanical equipment at the current moment and at the historical moment both exceed the preset health indicator failure threshold, then the target mechanical equipment is determined to be in a functional failure state at the current moment.

[0010] In an optional implementation, the method further includes using the target operating condition prediction module in the preset health benchmark model to predict the operating condition of the actual vibration data within the preset time window, and determining the predicted probability that the target mechanical equipment belongs to multiple preset operating conditions within the preset time window. Obtain the original historical vibration data of a preset mechanical device, wherein the preset mechanical device is a mechanical device of the same type as the target mechanical device, and the original historical vibration data is all vibration data of the preset mechanical device from the start of operation to the occurrence of degradation; Based on the original historical vibration data, multiple neurons corresponding to the initial health benchmark model are created, and the historical vibration data of the multiple neurons are obtained; the multiple neurons correspond to the multiple preset working conditions respectively; Based on the original historical vibration data, the initial working condition prediction module is trained to obtain the target working condition prediction module. Based on the historical vibration data of the multiple neurons, the initial prediction subnetworks corresponding to the multiple neurons are trained respectively to obtain the target prediction subnetworks for the multiple preset working conditions; Based on the target working condition prediction module and the target prediction subnet of the multiple preset working conditions, the preset health benchmark model is constructed.

[0011] In an optional implementation, the step of creating multiple neurons corresponding to the initial health baseline model based on the original historical vibration data includes: The original historical vibration data is divided into time windows to obtain historical vibration data for multiple time windows; The multiple neurons are created based on historical vibration data for each time window.

[0012] In an optional implementation, creating the plurality of neurons based on historical vibration data for each time window includes: The first neuron is created based on the historical vibration data of the first time window. Based on the historical vibration data of the second time window, determine whether the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window; If it is determined that the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window, then the first neuron is updated according to the historical vibration data of the second time window; If it is determined that the preset mechanical equipment does not meet the preset similarity conditions of the working condition corresponding to the first neuron under the second time window, then a second neuron is created based on the historical vibration data of the second time window, until the historical vibration data of each time window is judged, and the multiple neurons are obtained.

[0013] In an optional implementation, determining whether the preset mechanical equipment meets the preset similarity condition of the working condition corresponding to the first neuron under the second time window based on the historical vibration data of the second time window includes: Calculate the Mahalanobis distance corresponding to the second time window; If the Mahalanobis distance is greater than the preset Mahalanobis distance threshold, then it is determined that the preset mechanical equipment does not meet the preset similarity condition of the working condition corresponding to the first neuron in the second time window; If the Mahalanobis distance is less than or equal to a preset Mahalanobis distance threshold, then the preset mechanical equipment is determined to meet the preset similarity condition of the working condition corresponding to the first neuron under the second time window.

[0014] In an optional implementation, the step of training the initial prediction subnetworks corresponding to the plurality of neurons respectively based on the historical vibration data of the plurality of neurons to obtain the target prediction subnetworks for the plurality of preset working conditions includes: Based on the historical vibration data corresponding to the multiple neurons, the autoregressive features of the initial prediction subnet corresponding to the multiple neurons are determined; The state of the reserve pool of the initial prediction subnet corresponding to the multiple neurons is updated based on the input vector, input weight matrix, and internal weight matrix of the reserve pool of the initial prediction subnet corresponding to the multiple neurons. Based on the autoregressive features of the initial prediction subnetworks corresponding to the multiple neurons and the reserve pool state of the initial prediction subnetworks corresponding to the multiple neurons, the initial prediction subnetworks corresponding to the multiple neurons are trained to obtain the target prediction subnetworks for the multiple preset working conditions.

[0015] Secondly, embodiments of this application also provide a device for predicting the remaining life of mechanical equipment, comprising: The acquisition module is used to acquire the actual vibration data of the target mechanical equipment within a preset time window; the preset time window includes the current moment and a preset time period before the current moment. The prediction module is used to use the target working condition prediction module in the preset health benchmark model to predict the working condition of the actual vibration data in the preset time window and determine the prediction probability that the target mechanical equipment belongs to multiple preset working conditions in the preset time window. The prediction module is also used to use the target prediction subnet of the multiple preset working conditions in the preset health benchmark model to predict the actual vibration data of the preset time window and determine the predicted value of the vibration data of the target mechanical equipment for the multiple preset working conditions at the current time. The determination module is used to determine the health index of the target mechanical equipment at the current moment based on the predicted probabilities of the multiple preset working conditions, the predicted values ​​of the vibration data of the multiple preset working conditions, and the actual vibration data at the current moment. The determining module is also configured to determine the remaining service life of the target mechanical equipment at the current time based on the health indicators of the target mechanical equipment at the current time.

[0016] Thirdly, embodiments of this application also provide a computer device, including: a processor, a storage medium, and a bus, wherein the storage medium stores program instructions executable by the processor, and when the computer device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to perform the steps of the mechanical equipment remaining life prediction method as described in any of the first aspects.

[0017] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the mechanical equipment remaining life prediction method as described in any of the first aspects.

[0018] The beneficial effects of this application are: This application provides a method and apparatus for predicting the remaining lifespan of mechanical equipment. The method includes: acquiring actual vibration data of a target mechanical equipment within a preset time window; the preset time window includes the current moment and a preset time period prior to the current moment; using a target operating condition prediction module in a preset health benchmark model to predict the operating conditions of the actual vibration data within the preset time window, and determining the prediction probability that the target mechanical equipment belongs to multiple preset operating conditions within the preset time window; using a target prediction subnet of multiple preset operating conditions in the preset health benchmark model to predict the actual vibration data within the preset time window, and determining the predicted vibration data value of the target mechanical equipment for multiple preset operating conditions at the current moment; determining the health index of the target mechanical equipment at the current moment based on the prediction probability of multiple preset operating conditions, the predicted vibration data value of multiple preset operating conditions, and the actual vibration data at the current moment; and determining the remaining lifespan of the target mechanical equipment at the current moment based on the health index of the target mechanical equipment at the current moment. The method in this application acquires actual vibration data within a preset time window and combines it with a health benchmark model to achieve probability prediction of operating conditions, vibration data prediction, health index calculation, and remaining life calculation. The entire process is based on a probability density soft partitioning and online update mechanism, which not only avoids the risk of misjudgment in traditional hard partitioning of operating conditions, but also constructs a unified health index decoupled from operating conditions. This significantly improves the accuracy and robustness of equipment remaining life prediction under mixed operating conditions, providing a reliable basis for predictive maintenance. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 One of the flowcharts for a method to predict the remaining life of mechanical equipment provided in this application embodiment; Figure 2 A second schematic flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application embodiment; Figure 3 A third schematic flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application embodiment; Figure 4 A fourth schematic flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application embodiment; Figure 5 Fifth flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application embodiment; Figure 6A flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application is shown in Figure 6. Figure 7 The seventh flowchart illustrates a method for predicting the remaining life of mechanical equipment as provided in this application embodiment; Figure 8 This is the eighth flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application embodiment; Figure 9 A schematic diagram of the functional modules of a mechanical equipment remaining life prediction device provided in an embodiment of this application; Figure 10 This is a schematic diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0022] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] In the description of this application, it should be noted that if the terms "upper", "lower", etc. appear to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship that the product of this application is usually placed in, it is only for the convenience of describing this application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0024] Furthermore, the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Additionally, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] It should be noted that, where there is no conflict, the features in the embodiments of this application can be combined with each other.

[0026] The remaining life prediction method for mechanical equipment provided in this application will be explained in detail below with reference to the accompanying drawings and specific examples. The remaining life prediction method for mechanical equipment provided in this application can be implemented by a computer device pre-installed with a preset remaining life prediction processing algorithm or detection software for mechanical equipment, by running the algorithm or software. The computer device can be, for example, a server or a terminal, and the terminal can be a user computer. Figure 1 This application provides one of the flowcharts for a method to predict the remaining life of mechanical equipment; as shown in the embodiments. Figure 1 As shown, the method includes: S101. Obtain the actual vibration data of the target mechanical equipment within a preset time window.

[0027] The preset time window includes the current time and the preset time period before the current time.

[0028] In this embodiment, the target mechanical equipment can be bearings, gears, etc. of rotating machinery. An accelerometer mounted on the bearing housing collects vibration data in real time at the current time t and the preceding 0.8 seconds, with a preset time window length of 1 second, to form the actual vibration data.

[0029] S102. Using the target working condition prediction module in the preset health benchmark model, the actual vibration data of the preset time window is used to predict the working condition and determine the prediction probability that the target mechanical equipment belongs to multiple preset working conditions in the preset time window.

[0030] The pre-defined health baseline model employs a Modular AutoregRessive Echo State Network (M-ARESN), which comprises an input layer, a task decomposition layer, sub-network layers, and an ensemble output layer. The input layer receives the input data at the current time step; the task decomposition layer partitions the input data online based on probability density, assigning it to corresponding Gaussian components; the sub-network layers learn from the samples partitioned by the task decomposition layer to adapt to sub-tasks with different distributions; and finally, the ensemble output layer integrates and outputs the learning results from each sub-network layer.

[0031] The actual vibration data within the preset time window is input into the target working condition prediction module, i.e., the task decomposition layer, to calculate the probability density of the actual vibration data within the preset time window belonging to multiple preset working conditions. Then, the probability density of multiple preset working conditions is normalized to obtain the predicted probability of the target mechanical equipment belonging to multiple preset working conditions within the preset time window.

[0032] Specifically, the formula for calculating the probability density is:

[0033] in, This represents the actual vibration data within a preset time window. Let be the mean of the j-th neuron. Let the covariance matrix of the j-th neuron be represented. Each neuron corresponds to a preset working condition. Then the probability density of the actual vibration data in the preset time window belonging to multiple preset working conditions can be calculated.

[0034]

[0035] in, By normalizing the probability density of each of the multiple preset working conditions, the predicted probability of the target mechanical equipment belonging to multiple preset working conditions within a preset time window is obtained.

[0036] S103. Using a target prediction subnet of multiple preset working conditions in a preset health benchmark model, the actual vibration data of the preset time window is predicted to determine the predicted vibration data value of the target mechanical equipment for multiple preset working conditions at the current moment.

[0037] The actual vibration data of the preset time window is input into the target prediction subnet of multiple preset working conditions, i.e., the subnet layer. The target prediction subnet of multiple preset working conditions predicts the actual vibration data of the preset time window and determines the predicted value of the vibration data of the target mechanical equipment for multiple preset working conditions at the current moment.

[0038] S104. Based on the predicted probabilities of multiple preset working conditions, the predicted vibration data of multiple preset working conditions, and the actual vibration data at the current moment, determine the health indicators of the target mechanical equipment at the current moment.

[0039] S105. Determine the remaining service life of the target mechanical equipment at the current moment based on the health indicators of the target mechanical equipment at the current moment.

[0040] Specifically, the predicted probabilities of multiple preset working conditions, the predicted vibration data of multiple preset working conditions, and the actual vibration data at the current moment are processed to determine the health index of the target mechanical equipment at the current moment. Then, based on the health index of the target mechanical equipment at the current moment, the remaining service life of the target mechanical equipment at the current moment is determined.

[0041] In summary, this application provides a method and apparatus for predicting the remaining lifespan of mechanical equipment. The method includes: acquiring actual vibration data of a target mechanical equipment within a preset time window; the preset time window includes the current moment and a preset time period prior to the current moment; using a target operating condition prediction module in a preset health benchmark model to predict the operating condition of the actual vibration data within the preset time window, and determining the prediction probability that the target mechanical equipment belongs to multiple preset operating conditions within the preset time window; using a target prediction subnet for multiple preset operating conditions in the preset health benchmark model to predict the actual vibration data within the preset time window, and determining the predicted vibration data value of the target mechanical equipment for multiple preset operating conditions at the current moment; determining the health index of the target mechanical equipment at the current moment based on the prediction probability of multiple preset operating conditions, the predicted vibration data value of multiple preset operating conditions, and the actual vibration data at the current moment; and determining the remaining lifespan of the target mechanical equipment at the current moment based on the health index of the target mechanical equipment at the current moment. The method in this application acquires actual vibration data within a preset time window and combines it with a health benchmark model to achieve probability prediction of operating conditions, vibration data prediction, health index calculation, and remaining life calculation. The entire process is based on a probability density soft partitioning and online update mechanism, which not only avoids the risk of misjudgment in traditional hard partitioning of operating conditions, but also constructs a unified health index decoupled from operating conditions. This significantly improves the accuracy and robustness of equipment remaining life prediction under mixed operating conditions, providing a reliable basis for predictive maintenance.

[0042] This application also provides another possible implementation of the method for predicting the remaining life of mechanical equipment. Figure 2 This is a second schematic flowchart illustrating a method for predicting the remaining life of mechanical equipment, provided as an embodiment of this application. Figure 2 As shown, based on the predicted probabilities of multiple preset working conditions, the predicted vibration data of multiple preset working conditions, and the actual vibration data at the current moment, the health indicators of the target mechanical equipment at the current moment are determined, including: S201. Based on the predicted vibration data of multiple preset working conditions and the actual vibration data at the current moment, determine the prediction error of each preset working condition.

[0043] In this embodiment, the prediction error of the i-th preset working condition The calculation formula is:

[0044] in, This represents the predicted vibration data value for the i-th preset working condition. If the actual vibration data at the current moment is represented by the preset error calculation formula, the predicted vibration data value for each preset working condition and the actual vibration data at the current moment are calculated to obtain the prediction error for each preset working condition.

[0045] S202. Based on the prediction error of each preset working condition, the mean parameter and standard deviation parameter of each preset working condition, determine the standardized score corresponding to each preset working condition.

[0046] Specifically, to quantify the deviation of the current error from the healthy state, the prediction error of each preset working condition is compared with the error distribution established by its corresponding target prediction subnet under the healthy baseline state. The standardized score Z-Score for each preset working condition is calculated, where the standardized score for the i-th preset working condition is... The calculation formula is expressed as follows:

[0047] in, Let be the mean parameter of the i-th preset working condition. Let be the standard deviation parameter of the i-th preset working condition. Then, the standardized score calculation formula is used to calculate the prediction error, mean parameter and standard deviation parameter of each preset working condition to obtain the standardized score corresponding to each preset working condition.

[0048] The standardized score corresponding to each preset operating condition reflects the degree of difference between the current prediction error and the normal fluctuation level during a healthy period. When, it means that the performance of the target prediction subnet corresponding to the i-th preset working condition is no different from that during the healthy period; while when If the target prediction subnet corresponding to the i-th preset working condition shows obvious abnormality, then it can be considered that the performance of the target prediction subnet has become significantly abnormal.

[0049] S203. Determine the health indicators of the target mechanical equipment at the current moment based on the standardized score corresponding to each preset working condition and the predicted probability of each preset working condition.

[0050] Predict the probability of each preset working condition As weights, the standardized scores corresponding to each preset working condition That is, the health deviation of the target prediction subnet corresponding to each preset working condition is weighted and integrated to obtain the health index of the target mechanical equipment at the current moment.

[0051] Among them, the health indicators of the target mechanical equipment at the current moment The calculation formula is expressed as:

[0052] This allows us to obtain the health indicators of the target mechanical equipment at the current time, i.e., time t.

[0053] In the method provided in this application embodiment, the prediction error is first calculated by comparing the predicted value with the actual data, and then the standardized score is calculated by combining the mean and standard deviation of the error under the health benchmark. Finally, the health indicators are integrated by weighting the probability of the operating condition. This process not only quantifies the degree of deviation of the current operating state from the health state, but also realizes the fusion of health information under different operating conditions, ensuring the continuity and consistency of health indicators, and providing a stable and reliable measurement benchmark for subsequent remaining life prediction.

[0054] This application also provides another possible implementation of the method for predicting the remaining life of mechanical equipment. Figure 3 This is a third schematic flowchart illustrating a method for predicting the remaining life of mechanical equipment, provided as an embodiment of this application. Figure 3 As shown, based on the health indicators of the target machinery at the current moment, the remaining service life of the target machinery at the current moment is determined, including: S301. Determine the performance degradation rate of the target mechanical equipment at the current moment based on its health indicators at the current moment and its health indicators at historical moments.

[0055] In this embodiment, based on the health indicators of the target mechanical equipment at the current moment... Construct a time series using health indicators at historical moments. A sliding window-based linear regression model is used to fit the health indicator sequence within the most recent fixed time window in real time, so as to dynamically capture and quantify the current performance degradation rate of the device. .

[0056] S302. Determine the remaining service life of the target mechanical equipment at the current moment based on the performance degradation rate, the health indicators of the target mechanical equipment at the current moment, and the preset health indicator failure threshold.

[0057] Specifically, a predetermined health indicator failure threshold is set in advance based on the design margin or industry standards of the target mechanical equipment. According to the performance degradation rate Health indicators of the target mechanical equipment at the current moment and preset health indicator failure threshold Determine the remaining service life of the target mechanical equipment at the current moment, where the remaining service life... The calculation formula is:

[0058] Since performance degradation trajectory fitting is an online learning process, the performance degradation rate It will be continuously updated as health indicators change at new moments, so that the Remaining Useful Life (RUL) prediction can dynamically adapt to the latest performance degradation of the target machinery.

[0059] Optionally, if the health indicators of the target mechanical equipment at the current moment and the health indicators at historical moments both exceed the preset health indicator failure threshold, then the target mechanical equipment is determined to be in a functional failure state at the current moment.

[0060] Specifically, when the calculated health indicators of the target mechanical equipment at the current moment and at historical moments continuously exceed the preset health indicator failure threshold, the target mechanical equipment is determined to have entered a functional failure state.

[0061] In the method provided in this application embodiment, the performance degradation rate is dynamically captured by linear regression based on health indicators at historical and current times, and the remaining life is calculated by combining the preset failure threshold. The performance degradation rate is updated in real time with the new health indicators, which can adapt to the latest degradation performance of the equipment. This solves the problem of large fluctuations in the prediction results of traditional methods, improves the timeliness and accuracy of remaining life prediction, and helps to rationally plan the equipment maintenance cycle.

[0062] This application also provides another possible implementation of the method for predicting the remaining life of mechanical equipment. Figure 4 This is the fourth flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application. Figure 4 As shown, the method further includes using the target operating condition prediction module in the preset health benchmark model to predict the operating conditions of actual vibration data within a preset time window, and determining the prediction probability that the target mechanical equipment belongs to multiple preset operating conditions within the preset time window. S401. Obtain the original historical vibration data of the preset mechanical equipment.

[0063] Among them, the preset mechanical equipment is mechanical equipment of the same type as the target mechanical equipment, and the original historical vibration data is all vibration data from the start of operation of the preset mechanical equipment to the point before degradation occurs.

[0064] Specifically, if the target mechanical device is a gear, then the preset mechanical device is also a gear, and the original historical vibration data of the preset mechanical device is obtained. Based on the original historical vibration data of the preset mechanical device, an original time window is constructed.

[0065] S402. Based on the original historical vibration data, create multiple neurons corresponding to the initial health benchmark model, and obtain the historical vibration data of multiple neurons.

[0066] Among them, multiple neurons correspond to multiple preset working conditions.

[0067] Specifically, the original historical vibration data is processed to create multiple neurons.

[0068] Figure 5 This is the fifth flowchart illustrating a method for predicting the remaining life of mechanical equipment, provided as an embodiment of this application. Figure 5 As shown, step S402 specifically includes: S501. Divide the time window corresponding to the original historical vibration data to obtain historical vibration data for multiple time windows.

[0069] S502. Create multiple neurons based on historical vibration data for each time window.

[0070] The original historical vibration data is divided into time windows according to the size of a preset time window, thus obtaining historical vibration data for multiple time windows. , represented as: The historical vibration data for each time window is processed to create multiple neurons.

[0071] Figure 6 This is a sixth flowchart illustrating a method for predicting the remaining life of mechanical equipment, provided as an embodiment of this application. Figure 6 As shown, step S502 specifically includes: S601. Create the first neuron based on the historical vibration data of the first time window.

[0072] Specifically, features from different windows are extracted, and the features are assigned to different Gaussian components based on probability density, with each component corresponding to a working condition.

[0073] Before feature segmentation, the Gaussian component number of the initial health baseline model The value is 0, when the historical vibration data of the first window time window is... When inputting, with Create a mean centered at the center. The covariance matrix is The neuron is the first neuron, and the model is initialized with the following parameters:

[0074]

[0075] The subscript 1 represents the first neuron. Represented as an accumulator of the posterior probabilities of each neuron, storing the probabilities of the neurons so far. The cumulative posterior probability is then the accumulator of the posterior probability of the first neuron. =1, This represents the prior probability of the first neuron. This represents the number of times the first neuron is activated. It is a scalar parameter that controls the initial distribution range of the Gaussian components, and is initialized to 1. It is the covariance matrix of the first neuron. It is an identity matrix.

[0076] S602. Based on the historical vibration data of the second time window, determine whether the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window.

[0077] Among them, the preset similarity condition is used to characterize whether the working condition corresponding to the historical vibration data of the second time window is the same as the working condition corresponding to the historical vibration data of the first time window.

[0078] S603. If it is determined that the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron in the second time window, then the first neuron is updated according to the historical vibration data of the second time window.

[0079] Specifically, when inputting historical vibration data for the second time window... First, the probability density of historical vibration data in the first neuron is calculated for the second time window. and posterior probability .

[0080]

[0081]

[0082] If it is determined that the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron in the second time window, then the posterior probability corresponding to the historical vibration data of the second time window is... The value added to the accumulator of the posterior probability of the first neuron is represented as:

[0083] in This represents the value of the accumulator used to update the posterior probability of the first neuron. After calculating the cumulative posterior probability, the cumulative posterior probability of the first neuron in the decomposition layer is used as the prior for the next round, and the prior probability is reassigned through normalization.

[0084] by As the learning rate, update the mean and covariance matrix of the first neuron:

[0085]

[0086] S604. If it is determined that the preset mechanical equipment does not meet the preset similarity conditions of the working condition corresponding to the first neuron in the second time window, then create the second neuron based on the historical vibration data of the second time window, until the judgment of the historical vibration data of each time window is completed, and obtain multiple neurons.

[0087] If it is determined that the preset mechanical equipment does not meet the preset similarity conditions of the working condition corresponding to the first neuron in the second time window, then the second neuron is created with the historical vibration data of the second time window as the center. The creation method is the same as the creation method of the first neuron. Similarly, the historical vibration data of each time window is judged to determine whether the historical vibration data of each time window needs to create a neuron, and multiple neurons are obtained.

[0088] It should be noted that the initial health baseline model uses the cumulative posterior probability of judgment. Activation count and the number of samples The method for deciding whether to delete neurons This ensures the simplicity of the network structure. Specifically, when the target neuron is activated, in Number of activations within a sample time period The threshold was not reached. The number of times the target neuron is activated, or the number of times the target neuron is activated. Cumulative threshold However, the cumulative posterior probability Compared to the set threshold Small; means the target neuron If the neuron has a weak ability to fit the data, it would be a waste of resources to keep it and could lead to overfitting of the model. Therefore, the target neuron should be deleted.

[0089] S403. Based on the original historical vibration data, train the initial working condition prediction module to obtain the target working condition prediction module.

[0090] Specifically, the initial working condition prediction module is trained based on the original historical vibration data to determine the mean, covariance, and prior probability of each Gaussian component, forming the task decomposition layer, i.e., the target working condition prediction module.

[0091] S404. Based on the historical vibration data of multiple neurons, train the initial prediction subnetworks corresponding to multiple neurons respectively to obtain target prediction subnetworks for multiple preset working conditions.

[0092] S405. Based on the target working condition prediction module and multiple preset working condition target prediction subnets, construct a preset health benchmark model.

[0093] Specifically, based on the historical vibration data of multiple neurons, the initial prediction subnetworks corresponding to multiple neurons are trained to obtain multiple target prediction subnetworks for preset working conditions. Combined with the target working condition prediction module, a preset health benchmark model is trained to obtain the model.

[0094] In the method provided in this application embodiment, vibration data of the same type of equipment during the health stage are collected, neurons and prediction subnetworks adapted to different working conditions are created, and a complete health benchmark model is constructed. This health benchmark model integrates working condition identification and multi-subnetwork prediction functions, and the parameters are optimized through offline training, providing an accurate and reliable basic framework for subsequent online monitoring and life prediction, ensuring the scientificity and effectiveness of the entire prediction process.

[0095] This application also provides another possible implementation of the method for predicting the remaining life of mechanical equipment. Figure 7 This is the seventh flowchart illustrating a method for predicting the remaining life of mechanical equipment, provided as an embodiment of this application. Figure 7 As shown, based on the historical vibration data of the second time window, it is determined whether the preset mechanical equipment meets the preset similarity conditions of the corresponding working condition of the first neuron under the second time window, including: S701. Calculate the Mahalanobis distance corresponding to the second time window.

[0096] S702. If the Mahalanobis distance is greater than the preset Mahalanobis distance threshold, then it is determined that the preset mechanical equipment does not meet the preset similarity condition of the working condition corresponding to the first neuron in the second time window.

[0097] S703. If the Mahalanobis distance is less than or equal to the preset Mahalanobis distance threshold, then the preset mechanical equipment is determined to meet the preset similarity condition of the working condition corresponding to the first neuron in the second time window.

[0098] In this embodiment, the model employs an error-driven mechanism based on Mahalanobis distance to determine whether it is necessary to add neurons in each region to interpret the new data vector, i.e., the historical vibration data of the second time window. Assume the Mahalanobis distance between the historical vibration data of the second time window and the mean point of the first neuron in the task decomposition layer is... If the Mahalanobis distance If the distance is greater than the preset Mahalanobis distance threshold, it is determined that the preset mechanical equipment does not meet the preset similarity condition of the working condition corresponding to the first neuron in the second time window; if the Mahalanobis distance is greater than the preset similarity condition of the first neuron, it is determined that the preset mechanical equipment does not meet the preset similarity condition of the first neuron in the second time window. If the distance is less than or equal to the preset Mahalanobis distance threshold, then the preset mechanical equipment is determined to meet the preset similarity condition of the working condition corresponding to the first neuron in the second time window.

[0099] The formula for calculating the Mahalanobis distance for the second time window is as follows:

[0100] This can be understood as follows: The formula for calculating the Mahalanobis distance corresponding to the j-th neuron in the t-th time window is expressed as:

[0101] In the method provided in this application embodiment, the similarity of working conditions is judged by calculating Mahalanobis distance and comparing it with a preset Mahalanobis distance threshold. Mahalanobis distance can effectively take into account the data distribution characteristics and variable correlation. Compared with traditional distance measurement methods, it is more in line with the dynamic characteristics of vibration data, ensuring the accuracy of working condition judgment, providing a scientific basis for the reasonable updating or addition of neurons, and avoiding model inaccuracy caused by misjudgment of working conditions.

[0102] This application also provides another possible implementation of the method for predicting the remaining life of mechanical equipment. Figure 8 This is the eighth flowchart illustrating a method for predicting the remaining life of mechanical equipment provided in this application. Figure 8 As shown, based on the historical vibration data of multiple neurons, initial prediction subnetworks corresponding to multiple neurons are trained respectively to obtain target prediction subnetworks for multiple preset working conditions, including: S801. Based on the historical vibration data corresponding to multiple neurons, determine the autoregressive characteristics of the initial prediction subnet corresponding to multiple neurons.

[0103] In this embodiment, a corresponding target prediction subnet is trained for each preset working condition, and its task is to learn the dynamic change pattern of vibration data of healthy equipment under that working condition. To enhance the target prediction subnet's ability to model short-term dynamics of time series, an autoregressive (AR) feature of order p is introduced:

[0104] Where p is the order of the AR module, It contains the values ​​of the current time and the past p time steps.

[0105] S802. Update the state of the reserve pool of the initial prediction subnet corresponding to the multiple neurons based on the input vector, input weight matrix, and internal weight matrix of the reserve pool.

[0106] At time step t, the state of the reserve pool Update according to the following recursive formula:

[0107] in, , Leakage rate, Let be the input vector from the previous time step. For the input weight matrix, This is the internal weight matrix of the reserve pool.

[0108] S803. Based on the autoregressive features of the initial prediction subnetworks corresponding to multiple neurons and the reserve pool state of the initial prediction subnetworks corresponding to multiple neurons, train the initial prediction subnetworks corresponding to multiple neurons to obtain target prediction subnetworks for multiple preset working conditions.

[0109] Specifically, the output weights are updated online using the Recursive Least Squares (RLS) method. , First, construct the feature vector:

[0110] Where 1 is the bias term, updated using the RLS formula:

[0111]

[0112] in Let covariance matrix be the variance matrix. , , It is the identity matrix. For Kalman gain, The actual value at the current moment. The current prediction error is represented by the value of 'A'. Based on the autoregressive features of each subnet and the state of the reserve pool, the corresponding output weights are trained to obtain target prediction subnets for multiple preset working conditions. By introducing AR features, the reserve pool can explicitly capture the delay dependencies of the sequence, improving the short-term dynamic modeling capability.

[0113] For each prediction subnet, calculate its prediction error distribution on the health verification dataset, record the mean and standard deviation of the health status prediction error of each prediction subnet, and obtain target prediction subnets for multiple preset working conditions.

[0114] The formula for calculating the mean of the i-th prediction subnet is as follows:

[0115] The formula for calculating the standard deviation of the i-th prediction subnet is as follows:

[0116] In the method provided in this application embodiment, AR features are extracted to enhance the modeling ability of the prediction subnet for short-term dynamics of time series. The prediction subnet is trained through the state of the reserve pool, and online incremental updates of the output weights are realized, which improves the model training efficiency and fitting accuracy. This ensures that each subnet can accurately learn the vibration data pattern of the healthy equipment under the corresponding working conditions, and provides high-quality model support for subsequent vibration data prediction and health index calculation.

[0117] The following will continue to explain the mechanical equipment remaining life prediction device and computer equipment provided in any of the above embodiments of this application. The specific implementation process and the resulting technical effects are the same as those in the corresponding method embodiments. For the sake of brevity, the parts not mentioned in this embodiment can be referred to the corresponding content in the method embodiment.

[0118] Figure 9 This is a schematic diagram of the functional modules of a mechanical equipment remaining life prediction device provided in an embodiment of this application. Figure 9 As shown, the remaining life prediction device 100 for mechanical equipment includes: The acquisition module 110 is used to acquire the actual vibration data of the target mechanical equipment within a preset time window; the preset time window includes the current moment and a preset time period before the current moment. The prediction module 120 is used to use the target working condition prediction module in the preset health benchmark model to predict the working condition of the actual vibration data in the preset time window and determine the prediction probability of the target mechanical equipment belonging to multiple preset working conditions in the preset time window. The prediction module 120 is also used to use a target prediction subnet of multiple preset working conditions in a preset health benchmark model to predict the actual vibration data of a preset time window and determine the predicted value of the vibration data of the target mechanical equipment for multiple preset working conditions at the current moment. The determination module 130 is used to determine the health indicators of the target mechanical equipment at the current moment based on the predicted probabilities of multiple preset working conditions, the predicted values ​​of vibration data of multiple preset working conditions, and the actual vibration data at the current moment. The determination module 130 is also used to determine the remaining service life of the target mechanical equipment at the current moment based on the health indicators of the target mechanical equipment at the current moment.

[0119] Optionally, the determining module 130 is further configured to determine the prediction error of each preset working condition based on the predicted vibration data of multiple preset working conditions and the actual vibration data at the current moment; determine the standardized score corresponding to each preset working condition based on the prediction error of each preset working condition, the mean parameter and the standard deviation parameter of each preset working condition; and determine the health index of the target mechanical equipment at the current moment based on the standardized score corresponding to each preset working condition and the prediction probability of each preset working condition.

[0120] Optionally, the determining module 130 is further configured to determine the performance degradation rate of the target mechanical equipment at the current moment based on the health indicators of the target mechanical equipment at the current moment and the health indicators at historical moments; and to determine the remaining service life of the target mechanical equipment at the current moment based on the performance degradation rate, the health indicators of the target mechanical equipment at the current moment, and the preset health indicator failure threshold.

[0121] Optionally, the determining module 130 is further configured to determine that the target mechanical equipment is in a functional failure state at the current moment if both the health indicators of the target mechanical equipment at the current moment and the health indicators at historical moments exceed the preset health indicator failure threshold.

[0122] Optionally, the device further includes: The acquisition module 110 is used to acquire the original historical vibration data of the preset mechanical equipment. The preset mechanical equipment is a type of mechanical equipment that is the same as the target mechanical equipment. The original historical vibration data is all the vibration data of the preset mechanical equipment from the start of operation to the point of degradation. The module is used to create multiple neurons corresponding to the initial health benchmark model based on the original historical vibration data, and to obtain the historical vibration data of the multiple neurons; the multiple neurons correspond to multiple preset working conditions respectively; The training module is used to train the initial working condition prediction module based on the original historical vibration data to obtain the target working condition prediction module. The training module is also used to train the initial prediction subnets corresponding to multiple neurons based on the historical vibration data of multiple neurons, so as to obtain target prediction subnets for multiple preset working conditions. The construction module is used to build a preset health baseline model based on the target working condition prediction module and multiple preset working condition target prediction subnets.

[0123] Optionally, the module is also used to divide the time window corresponding to the original historical vibration data to obtain historical vibration data for multiple time windows; and to create multiple neurons based on the historical vibration data of each time window.

[0124] Optionally, the creation module is further configured to: create a first neuron based on historical vibration data from the first time window; determine whether the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window based on historical vibration data from the second time window; if the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window, then update the first neuron based on historical vibration data from the second time window; if the preset mechanical equipment does not meet the preset similarity conditions of the working condition corresponding to the first neuron under the second time window, then create a second neuron based on historical vibration data from the second time window, until the historical vibration data of each time window is judged, resulting in multiple neurons.

[0125] Optionally, module 130 is used to calculate the Mahalanobis distance corresponding to the second time window; if the Mahalanobis distance is greater than a preset Mahalanobis distance threshold, it is determined that the preset mechanical equipment does not meet the preset similarity condition of the working condition corresponding to the first neuron under the second time window; if the Mahalanobis distance is less than or equal to the preset Mahalanobis distance threshold, it is determined that the preset mechanical equipment meets the preset similarity condition of the working condition corresponding to the first neuron under the second time window.

[0126] Optionally, the training module is used to determine the autoregressive features of the initial prediction subnet corresponding to the multiple neurons based on the historical vibration data corresponding to the multiple neurons; update the reserve pool state of the initial prediction subnet corresponding to the multiple neurons based on the input vector, input weight matrix, and internal weight matrix of the reserve pool; and train the initial prediction subnet corresponding to the multiple neurons based on the autoregressive features and the reserve pool state of the initial prediction subnet corresponding to the multiple neurons to obtain target prediction subnets for multiple preset working conditions.

[0127] The above-described device is used to execute the method provided in the foregoing embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

[0128] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more microprocessors, or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).

[0129] Figure 10 This is a schematic diagram of a computer device provided in an embodiment of this application. This computer device can be used for predicting the remaining lifespan of mechanical equipment. Figure 10 As shown, the computer device includes: a processor 210, a storage medium 220, and a bus 230.

[0130] Storage medium 220 stores machine-readable instructions executable by processor 210. When the computer device is running, processor 210 communicates with storage medium 220 via bus 230, and processor 210 executes the machine-readable instructions to perform the steps of the above method embodiment. The specific implementation and technical effects are similar, and will not be described again here.

[0131] Optionally, this application also provides a storage medium 220, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the above-described method embodiments. The specific implementation and technical effects are similar, and will not be repeated here.

[0132] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0133] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0134] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0135] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting the remaining life of mechanical equipment, characterized in that, include: Acquire the actual vibration data of the target mechanical equipment within a preset time window; The preset time window includes the current moment and a preset time period preceding the current moment. Using the target working condition prediction module in the preset health benchmark model, the actual vibration data of the preset time window is used to predict the working condition and determine the predicted probability that the target mechanical equipment belongs to multiple preset working conditions in the preset time window. Using the target prediction subnet of the multiple preset working conditions in the preset health benchmark model, the actual vibration data of the preset time window is predicted to determine the predicted value of the vibration data of the target mechanical equipment for the multiple preset working conditions at the current time. Based on the predicted probabilities of the multiple preset working conditions, the predicted vibration data of the multiple preset working conditions, and the actual vibration data at the current moment, the health index of the target mechanical equipment at the current moment is determined. Based on the health indicators of the target mechanical equipment at the current moment, determine the remaining service life of the target mechanical equipment at the current moment.

2. The method according to claim 1, characterized in that, The step of determining the health index of the target mechanical equipment at the current moment based on the predicted probabilities of the multiple preset working conditions, the predicted vibration data values ​​of the multiple preset working conditions, and the actual vibration data at the current moment includes: Based on the predicted vibration data of the multiple preset working conditions and the actual vibration data at the current moment, the prediction error of each preset working condition is determined. Based on the prediction error of each preset working condition, the mean parameter and standard deviation parameter of each preset working condition, the standardized score corresponding to each preset working condition is determined. Based on the standardized score corresponding to each preset working condition and the predicted probability of each preset working condition, the health index of the target mechanical equipment at the current moment is determined.

3. The method according to claim 1, characterized in that, The step of determining the remaining service life of the target mechanical equipment at the current moment based on the health indicators of the target mechanical equipment at the current moment includes: Based on the health indicators of the target mechanical equipment at the current moment and the health indicators at historical moments, determine the performance degradation rate of the target mechanical equipment at the current moment; The remaining service life of the target mechanical equipment at the current moment is determined based on the performance degradation rate, the health index of the target mechanical equipment at the current moment, and the preset health index failure threshold.

4. The method according to claim 3, characterized in that, The method further includes: If the health indicators of the target mechanical equipment at the current moment and at the historical moment both exceed the preset health indicator failure threshold, then the target mechanical equipment is determined to be in a functional failure state at the current moment.

5. The method according to claim 1, characterized in that, The method further includes: using the target operating condition prediction module in the preset health benchmark model to predict the operating condition of the actual vibration data within the preset time window, and determining the prediction probability that the target mechanical equipment belongs to multiple preset operating conditions within the preset time window. Obtain the original historical vibration data of a preset mechanical device, wherein the preset mechanical device is a mechanical device of the same type as the target mechanical device, and the original historical vibration data is all vibration data of the preset mechanical device from the start of operation to the occurrence of degradation; Based on the original historical vibration data, multiple neurons corresponding to the initial health benchmark model are created, and the historical vibration data of the multiple neurons are obtained; the multiple neurons correspond to the multiple preset working conditions respectively; Based on the original historical vibration data, the initial working condition prediction module is trained to obtain the target working condition prediction module. Based on the historical vibration data of the multiple neurons, the initial prediction subnetworks corresponding to the multiple neurons are trained respectively to obtain the target prediction subnetworks for the multiple preset working conditions; Based on the target working condition prediction module and the target prediction subnet of the multiple preset working conditions, the preset health benchmark model is constructed.

6. The method according to claim 5, characterized in that, The process of creating multiple neurons corresponding to the initial health baseline model based on the original historical vibration data includes: The original historical vibration data is divided into time windows to obtain historical vibration data for multiple time windows; The multiple neurons are created based on historical vibration data for each time window.

7. The method according to claim 6, characterized in that, The process of creating the multiple neurons based on historical vibration data for each time window includes: The first neuron is created based on the historical vibration data of the first time window. Based on the historical vibration data of the second time window, determine whether the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window; If it is determined that the preset mechanical equipment meets the preset similarity conditions of the working condition corresponding to the first neuron under the second time window, then the first neuron is updated according to the historical vibration data of the second time window; If it is determined that the preset mechanical equipment does not meet the preset similarity conditions of the working condition corresponding to the first neuron under the second time window, then a second neuron is created based on the historical vibration data of the second time window, until the historical vibration data of each time window is judged, and the multiple neurons are obtained.

8. The method according to claim 7, characterized in that, The step of determining whether the preset mechanical equipment meets the preset similarity condition of the working condition corresponding to the first neuron under the second time window based on the historical vibration data of the second time window includes: Calculate the Mahalanobis distance corresponding to the second time window; If the Mahalanobis distance is greater than the preset Mahalanobis distance threshold, then it is determined that the preset mechanical equipment does not meet the preset similarity condition of the working condition corresponding to the first neuron in the second time window; If the Mahalanobis distance is less than or equal to a preset Mahalanobis distance threshold, then the preset mechanical equipment is determined to meet the preset similarity condition of the working condition corresponding to the first neuron under the second time window.

9. The method according to claim 5, characterized in that, The step of training the initial prediction subnetworks corresponding to the multiple neurons based on their historical vibration data to obtain the target prediction subnetworks for the multiple preset working conditions includes: Based on the historical vibration data corresponding to the multiple neurons, the autoregressive features of the initial prediction subnet corresponding to the multiple neurons are determined; The state of the reserve pool of the initial prediction subnet corresponding to the multiple neurons is updated based on the input vector, input weight matrix, and internal weight matrix of the reserve pool of the initial prediction subnet corresponding to the multiple neurons. Based on the autoregressive features of the initial prediction subnetworks corresponding to the multiple neurons and the reserve pool state of the initial prediction subnetworks corresponding to the multiple neurons, the initial prediction subnetworks corresponding to the multiple neurons are trained to obtain the target prediction subnetworks for the multiple preset working conditions.

10. A computer device, characterized in that, include: The system includes a processor, a storage medium, and a bus, wherein the storage medium stores program instructions executable by the processor, and when the computer device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to perform the steps of the mechanical equipment remaining life prediction method as described in any one of claims 1 to 9.

Images