Equipment health management method and device based on small sample data and adaptive migration

Abstract

The invention discloses an equipment health management method and device based on small sample data and adaptive migration, and relates to the technical field of equipment health management. Health indexes are constructed for sensor data through a BLCAE model and a Pearson's correlation coefficient; secondly, judging the initial degradation starting time of the equipment through a degradation stage detection method, and extracting degradation stage data; and finally, migration is carried out among the same degradation category, so that degradation information can be effectively shared among similar fields or tasks, and the precision and stability of the model are improved. The method comprises the following steps: constructing a health index of target equipment based on sensor signal data, and detecting whether the target equipment is degraded or not according to the health index; if it is determined that the target equipment degenerates according to the health index detection, determining a degeneration type of the target equipment and extracting degeneration stage data; and inputting the degradation stage data into a health state prediction model corresponding to the degradation type to obtain a health state prediction result output by the health state prediction model.

Description

Technical Field

[0001] The present application relates to the technical field of equipment health management, and in particular to an equipment health management method and apparatus based on small sample data and adaptive migration. Background Art

[0002] With the development of equipment management mode, deep learning has received extensive attention and research in the field of equipment health management. Since health management is the key to ensuring efficient and sustainable service of equipment in the future, potential failures can be predicted in advance by predicting trends, behavior patterns and correlations through statistical or machine learning models, thereby improving the decision-making process of maintenance activities, avoiding downtime, and reducing personal injury and property losses caused by sudden failures. Therefore, real-time health management of equipment, especially in small sample cases, has positive practical significance. However, the small sample problem is one of the main reasons that hinder the widespread application of deep learning in this field.

[0003] In related technologies, some scholars have applied transfer learning to equipment health management under small samples, sharing and migrating information between similar fields or tasks. Feature-based deep transfer learning methods reduce the distribution deviation between the source domain and the target domain by minimizing the maximum average difference between the two to solve the small sample problem, data distribution difference problem and personalization problem.

[0004] In the process of implementing this application, the applicant found that the related technology has at least the following problems:

[0005] Affected by the working environment, working conditions and individual differences, the degradation process and probability distribution of the same equipment or parts may be significantly different. When the distribution difference in the field is too large and the data quality is too poor, forced transmission may lead to negative migration, and the model performance will be worse. Especially in the case of small samples, it is difficult for the model to avoid the interference of extreme data on the overall data. Moreover, the degradation stage judgment method using a single recognition strategy and a fixed alarm threshold cannot adapt to individual differences, and it is difficult to cope with the interference of random noise in the data. It is impossible to accurately extract the degradation stage data, which will not only waste computing resources and reduce the efficiency of model training, but also affect the accuracy of equipment health management and lead to misjudgment. Summary of the invention

[0006] In view of this, the present application provides a method and device for equipment health management based on small sample data and adaptive migration, the main purpose of which is to solve the following problems: affected by the working environment, working conditions and individual differences, the degradation process and probability distribution of the same equipment or parts may be significantly different. Especially in the case of small samples, it is difficult for the model to avoid the interference of extreme data on the overall data. Moreover, the degradation stage judgment method using a single recognition strategy and a fixed alarm threshold cannot adapt to individual differences, is difficult to cope with the interference of random noise in the data, and cannot accurately extract the degradation stage data, which will not only waste computing resources and reduce the efficiency of model training, but also affect the accuracy of equipment health management and lead to misjudgment.

[0007] According to a first aspect of the present application, a device health management method based on small sample data and adaptive migration is provided, the method comprising:

[0008] Acquire sensor signal data of a target device, and construct a health indicator of the target device based on the sensor signal data;

[0009] Detecting whether the target device is degraded according to the health indicator;

[0010] If the target device is determined to be degraded according to the health indicator detection, then determining the degradation type of the target device, and extracting degradation stage data from the sensor signal data based on the degradation stage detection method and the health indicator;

[0011] A health status prediction model corresponding to the degradation type is obtained, and the degradation stage data is input into the health status prediction model to obtain a health status prediction result output by the health status prediction model.

[0012] According to a second aspect of the present application, a device health management apparatus based on small sample data and adaptive migration is provided, the apparatus comprising:

[0013] A construction module, used to obtain sensor signal data of a target device, and construct a health indicator of the target device based on the sensor signal data;

[0014] A detection module, used to detect whether the target device is degraded according to the health indicator;

[0015] an extraction module, configured to determine the degradation type of the target device if it is determined that the target device is degraded according to the health indicator detection, and extract degradation stage data from the sensor signal data based on the degradation stage detection method and the health indicator;

[0016] The prediction module is used to obtain a health status prediction model corresponding to the degradation type, input the degradation stage data into the health status prediction model, and obtain a health status prediction result output by the health status prediction model.

[0017] Through the above-mentioned ending scheme, the present application provides a device health management method and device based on small sample data and adaptive migration. The present application obtains sensor signal data of the target device, constructs the health index of the target device based on the sensor signal data, and detects whether the target device has been degraded according to the health index; if it is determined that the target device has been degraded according to the health index detection, the degradation type of the target device is determined, and the degradation stage data is extracted from the sensor signal data based on the degradation stage detection method and the health index; a health status prediction model corresponding to the degradation type is obtained, and the degradation stage data is input into the health status prediction model to obtain a health status prediction result output by the health status prediction model. Firstly, a reasonable health index is constructed for sensor data through BLCAE (bidirectional long short-term memory convolutional neural network model) and Pearson correlation coefficient (Pearson correlation coefficient); secondly, the degradation stage detection method based on OSW-PI-ACED (overlapping sliding window-Pettitt mutation point, inflection point extraction-adaptive continuous excess detection verification) is used to determine the initial degradation start time of the equipment and extract the degradation stage data of each data; finally, migration is performed between the same degradation categories, so that the degradation information can be effectively shared between similar fields or tasks, while avoiding negative migration and improving the reusability of historical knowledge and information. By performing necessary processing on the limited data collected by the sensor, the data is classified to reduce the feature distribution difference between the source domain and the target domain, and domain adaptive migration is performed between the same categories, so that the degradation information can be shared between similar fields or tasks, avoiding the occurrence of negative migration, and accurately extracting the degradation stage data of the equipment, reducing the interference of irrelevant information on the prediction of the health status of the equipment. Even in the case of small samples, it is possible to capture degradation information from a small amount of data, reduce cross-domain data deviations, improve the accuracy and stability of the model, and achieve health management of the equipment.

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0020] Figure 1 A schematic diagram of a method flow of device health management based on small sample data and adaptive migration provided in an embodiment of the present application is shown;

[0021] Figure 2A A schematic diagram of another method flow for device health management based on small sample data and adaptive migration provided in an embodiment of the present application is shown;

[0022] Figure 2B A schematic diagram of a health status prediction process provided by an embodiment of the present application is shown;

[0023] Figure 3A A schematic diagram of the structure of device health management based on small sample data and adaptive migration provided by an embodiment of the present application is shown;

[0024] Figure 3B A schematic diagram of another structure of device health management based on small sample data and adaptive migration provided by an embodiment of the present application is shown;

[0025] Figure 4 A schematic diagram of the device structure of a computer device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0027] The present application embodiment provides a device health management method based on small sample data and adaptive migration, such as Figure 1 As shown, the method includes:

[0028] 101. Obtain sensor signal data of the target device, and construct a health indicator of the target device based on the sensor signal data.

[0029] The health stage data provides little information for predicting the health status of the equipment. The current degradation stage judgment method that uses a single recognition strategy and a fixed alarm threshold cannot adapt to individual differences and is difficult to cope with the interference of random noise in the data. If the degradation stage data cannot be accurately extracted, it will not only waste computing resources and reduce the efficiency of model training, but also affect the accuracy of equipment health management and lead to misjudgment. The feature-based deep transfer learning method reduces the distribution deviation between the source domain and the target domain by minimizing the maximum average difference between the two. However, if the domain distribution difference is too large and the data quality is too poor, it may lead to negative transfer, and the model performance will be worse. Especially in the case of small samples, it is difficult for the model to cope with the interference of extreme data on the overall data.

[0030] To solve this problem, this application proposes a device health management method based on small sample data and adaptive migration. First, a reasonable health indicator is constructed for sensor data through BLCAE (bidirectional long short-term memory convolutional neural network model) and Pearson correlation coefficient (Pearson correlation coefficient); secondly, the degradation stage detection method based on OSW-PI-ACED (overlapping sliding window-Pettitt mutation point, inflection point extraction-adaptive continuous exceedance detection verification) is used to determine the initial degradation start time of the equipment, and the degradation stage data of each data is extracted; finally, migration is performed between the same degradation categories, so that the degradation information can be effectively shared between similar fields or tasks, thereby improving the reusability of historical knowledge and information while avoiding negative migration. The executor of this application may be a device health management system, which relies on the computing power of the server to provide services to users. The server may be an independent server or may provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), as well as servers for basic cloud computing such as big data and artificial intelligence platforms, so that the device health management system can quickly, accurately and stably realize device health management.

[0031] In an embodiment of the present application, the equipment health management system obtains the sensor signal data of the target equipment and constructs the health index of the target equipment based on the sensor signal data. Among them, the sensor signal data can be any sensor signal data containing equipment degradation information in actual applications, such as vibration data, temperature data, humidity data, etc., which can be obtained from sensors installed on the equipment itself. It can also be sensor signal data of equipment under different working conditions. In this case, the sensor signal data refers to the data containing equipment degradation information from the first start of operation to failure of the equipment, such as: full-life vibration signal data, full-life temperature signal data, etc.

[0032] 102. Detect whether the target device has degraded based on the health indicators.

[0033] In an embodiment of the present application, the equipment health management system detects whether the target equipment has degraded based on the health index. Taking the change in vibration amplitude during bearing degradation as an example, in its full-life vibration data, the vibration amplitude is in a fluctuating state, and is not linearly related to time. Moreover, the change in vibration amplitude is a process from slight to significant increase. When the amplitude change shows an upward trend, it indicates that the bearing has begun to degrade. The value range of the health index is from 1 to 0, which can adapt to sensor data of different components and different working conditions, and thus more clearly determine whether the target equipment has degraded. In the healthy stage, the health index is basically between [0.95, 1]. In the degradation stage, the value of the health index gradually decreases to 0. In simple terms, for example, [2, 2.1, 1.99, 1.98, 2…2.5, 2.7, 2.6, 2.8…3…20] is regarded as the full-life vibration data, and [1, 0.99, 0.99, 1, 0.97…0.99, 0.95, 0.92, …0.2…0] is regarded as the health index corresponding to the full-life vibration data. If the health index constructed during the detection is [0.99, 1, 0.99, 0.98…0.99], it means that the current bearing is in a healthy state and no health state prediction is required; if the health index constructed during the detection appears [0.98, 0.99, 1, 0.97, 0.8…0.2…0.01], that is, the health index as a whole begins to gradually decrease with time, then it means that the bearing has begun to degrade and a health state prediction is required.

[0034] 103. If it is determined that the target device has degraded according to the health indicator detection, the degradation type of the target device is determined, and the degradation stage data is extracted from the sensor signal data based on the degradation stage detection method and the health indicator.

[0035] In the embodiment of the present application, if the equipment health management system determines that the target equipment has degraded based on the health index detection, the degradation type of the target equipment is determined, and the degradation stage data is extracted from the sensor signal data based on the degradation stage detection method and the health index, wherein the degradation type includes a rapid degradation type and a slow degradation type. When it is determined that the bearing has begun to degrade, it is necessary to determine the degradation type of the test bearing, and then extract the degradation stage data and put it into the model of the corresponding degradation type for health status prediction.

[0036] 104. Obtain a health status prediction model corresponding to the degradation type, input the degradation stage data into the health status prediction model, and obtain a health status prediction result output by the health status prediction model.

[0037] In the embodiment of the present application, the equipment health management system obtains the health status prediction model corresponding to the degradation type, inputs the degradation stage data into the health status prediction model, and obtains the health status prediction result output by the health status prediction model. The present application uses two types of degradation type data to train two models, constructs health indicators for the equipment through sensor signals, judges the current equipment health status based on the health indicators, extracts data of the equipment in the degradation stage, completes data classification at the same time, and migrates between the same degradation categories to reduce data distribution differences and avoid the occurrence of negative migration, thereby realizing equipment health management under small samples.

[0038] The method provided in the embodiment of the present application obtains the sensor signal data of the target device, constructs the health index of the target device based on the sensor signal data, and detects whether the target device has degraded according to the health index; if the target device is determined to be degraded according to the health index detection, the degradation type of the target device is determined, and the degradation stage data is extracted from the sensor signal data based on the degradation stage detection method and the health index; the health state prediction model corresponding to the degradation type is obtained, and the degradation stage data is input into the health state prediction model to obtain the health state prediction result output by the health state prediction model. First, a reasonable health index is constructed for the sensor data through BLCAE (bidirectional long short-term memory convolutional neural network model) and Pearson correlation coefficient (Pearson correlation coefficient); secondly, the initial degradation start time of the device is determined by the degradation stage detection method of OSW-PI-ACED (based on overlapping sliding window-Pettitt mutation point, inflection point extraction-adaptive continuous exceedance detection verification), and the degradation stage data of each data is extracted; finally, migration is performed between the same degradation categories, so that the degradation information can be effectively shared between similar fields or tasks, while avoiding negative migration and improving the reusability of historical knowledge and information. By performing necessary processing on the limited data collected by the sensor, the data is classified to reduce the feature distribution differences between the source domain and the target domain, and domain adaptive migration is performed between the same categories, so that degradation information can be shared between similar fields or tasks to avoid the occurrence of negative migration, and the degradation stage data of the equipment can be accurately extracted to reduce the interference of irrelevant information on the prediction of the health status of the equipment. Even in the case of small samples, degradation information can be captured from a small amount of data, reducing data cross-domain deviation, improving the accuracy and stability of the model, and realizing health management of the equipment.

[0039] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides another device health management method based on small sample data and adaptive migration, such as Figure 2A As shown, the method includes:

[0040] 201. Obtain sensor signal sample data, obtain sample data with labels from the sensor signal sample data as source domain data, and obtain sample data without labels from the sensor signal sample data as target domain data.

[0041] In the embodiment of the present application, the equipment health management system first obtains sensor signal sample data, wherein the sensor signal sample data is the equipment timing degradation data under different working conditions. Since the vibration of the equipment will change during the degradation process, the equipment timing degradation data in the embodiment of the present application mainly refers to vibration data. For example, the sampling time is set to 2s and the sampling interval is 1min, and the vibration amplitude, acceleration and other signals of the equipment are collected through the sensor on the equipment to obtain the equipment timing degradation data. The equipment timing degradation data includes the vibration signal data of the entire process from health to failure of the equipment. The data collected in the embodiment of the present application is the vibration data in the bearing degradation process, including PHM 2012Dataset and XJTU-SY Dataset. Taking XJTU-SY as an example, the data set contains three working conditions, and each working condition has five full-life vibration data of bearings from health to failure, and PHM 2012 is similar. Then, the equipment health management system obtains the sample data with labels in the sensor signal sample data as the source domain data, and obtains the sample data without labels in the sensor signal sample data as the target domain data. By performing necessary processing on the limited data collected by the sensor, the data is classified to reduce the feature distribution differences between the source domain and the target domain, and domain adaptive migration is performed between the same categories, so that the degraded information can be shared between similar domains or tasks to avoid the occurrence of negative transfer.

[0042] It should be noted that in practical applications, the sensor signal sample data includes both labeled sample data and unlabeled sample data. However, if the data in the two data sets used for training and verification are unlabeled, it is necessary to extract several groups of data from all the data and label them as the source domain. The data here refers to the sensor signal sample data. When using these two unlabeled data sets for model training and verification, in order to simplify the process, some data is extracted from the two categories after the degradation process classification is completed, and then labeled as the source domain.

[0043] 202. Divide the target domain data into a training data set and a test data set according to a preset ratio, and use the source domain data and the training data set as target training data sets.

[0044] In an embodiment of the present application, the equipment health management system divides the target domain data into a training data set and a test data set according to a preset ratio, and uses the source domain data and the training data set as the target training data set, wherein the preset ratio can be set to 80% as the training set and 20% as the test set.

[0045] 203. Obtain a pre-built bidirectional long short-term memory convolutional neural network model and a Pearson correlation coefficient, and use the bidirectional long short-term memory convolutional neural network model and the Pearson correlation coefficient to process the target training data set to obtain a first health indicator.

[0046] In an embodiment of the present application, the equipment health management system obtains a pre-constructed bidirectional long short-term memory convolutional neural network model and a Pearson correlation coefficient, and uses the bidirectional long short-term memory convolutional neural network model and the Pearson correlation coefficient to process the target training data set to obtain a first health indicator. Among them, the BLCAE model is a Bi-LSTM-CNN autoencoder, which is used to extract features containing degradation information in the equipment timing degradation data, and the features output by the BLCAE model are a three-dimensional matrix, which can not only effectively filter out the random noise in the original data, but also make the constructed health indicators more representative. By extracting features through the BLCAE model, and then using the Pearson correlation coefficient to obtain the correlation between features, the health indicator is obtained, which can adapt to sensor data of different components and different working conditions, thereby improving the stability and extensiveness of the health status prediction model.

[0047] 204. According to the degradation type, the target training data set is divided into a rapid degradation data set and a slow degradation data set by using the first health indicator, and the rapid degradation stage data is extracted from the rapid degradation data set and the slow degradation stage data is extracted from the slow degradation data set by using the degradation stage detection method.

[0048] The health index can describe the degradation process of the equipment. Some equipment takes a long time from the beginning of degradation to failure, which can be called slow degradation; while some only take a short time, which can be called rapid degradation. According to the constructed health index, the data of the same degradation type are divided into one category, which can reduce the difference in data distribution and avoid the occurrence of negative migration. In the embodiment of the present application, the equipment health management system divides the first health indicator into the health indicator corresponding to the rapid degradation type and the health indicator corresponding to the slow degradation type according to the degradation type, and then divides the target training data set into a rapid degradation data set and a slow degradation data set based on the health indicator corresponding to the rapid degradation type and the health indicator corresponding to the slow degradation type. Among them, the degradation type can be simply understood as the ratio of the time t from the beginning of degradation to failure of the equipment to the time T from the beginning of use to failure of the equipment, that is, t / T, which is divided into two categories or n categories according to the ratio and the size of the data. In the embodiment of the present application, the degradation type is divided into rapid degradation (t / T is relatively small) and slow degradation (t / T is relatively large).

[0049] Next, the equipment health management system uses the degradation stage detection method to extract rapid degradation stage data from the rapid degradation data set, and uses the degradation stage detection method to extract slow degradation stage data from the slow degradation data set. Among them, the degradation stage detection method is a degradation stage detection method based on overlapping sliding windows-Pettitt mutation points, inflection point extraction-adaptive continuous exceedance detection verification, which is used to determine the initial degradation start time of the equipment and extract degradation stage data.

[0050] Specifically, the equipment health management system uses the 2sigma principle to extract the target health indicator from the health indicators corresponding to the rapid degradation type, and uses the target health indicator to determine the detection threshold. Then, the equipment health management system uses an overlapping sliding window to slide over the target health indicator. During the sliding process, the mutation point of the target health indicator is extracted from the multiple data points included in the overlapping sliding window based on the mutation point detection algorithm, and multiple mutation points corresponding to each overlapping sliding window are obtained.

[0051] For each overlapping sliding window, the equipment health management system uses multiple data points included in the overlapping sliding window to determine multiple inflection points. Specifically, for each data point, a target data point is determined among multiple data points based on the data point, wherein the target data point is the third data point behind the data point. Then, the slope value of the data point and the target data point is calculated; if the slope value is less than the preset threshold, the data point is used as an inflection point, wherein the preset threshold is -0.002. Next, the equipment health management system sorts the multiple mutation points and the multiple inflection points in chronological order to obtain a time series. The equipment health management system extracts multiple target time points in the time series according to the principle of continuous excess detection and verification. Specifically, according to the principle of continuous excess detection and verification, detection starts from the first time point of the time series; if the i-th time point in the time series is greater than the detection threshold, the i+1-th time point continues to be detected until the i+j-th time point is detected. The calculation formula of the parameter j is the following formula 1:

[0052] Formula 1:

[0053] Where n is the number of time points in the time series. When j-1 time points are detected to be greater than the detection threshold, the i-th time point is taken as the target time point. Multiple target time points can be obtained by detecting until the last time point of the time series.

[0054] Subsequently, the equipment health management system sorts multiple target time points in chronological order to obtain a sorting result, and reads the first target time point in the sorting result as the initial degradation time. Finally, the equipment health management system extracts the rapid degradation stage data from the rapid degradation data set according to the initial degradation time.

[0055] Taking a set of full-life data as an example, the health index is first extracted using overlapping sliding windows to avoid misjudgment of the data at the end of the window, where the window size and overlap size are set to 100 and 15 respectively. Then, the Pettitt (mutation point detection) method is used to extract the mutation points in the window, and the slope between each point in each window and the third point after it is calculated. If the slope is less than the set threshold, i.e. -0.002, the point is considered to be an inflection point and may be a degradation point. Subsequently, the detected mutation points and inflection points are placed in the same list in chronological order. Finally, according to the 2sigma principle, the first 10% of the health index is used to construct the detection threshold, and according to the continuous exceedance detection verification principle, first determine whether the i-th point in the list exceeds the detection threshold. If it exceeds the detection threshold, determine whether j-1 points in the i+1, i+2,...i+j points exceed the threshold, where j is determined by the total number of mutation points in the list n. If j-1 points exceed the threshold, it means that the i-th point meets the requirements. Then all the points that meet the requirements are put into a new list in chronological order, and the time corresponding to the first value in the new list is taken as the initial degradation time. The original data after the initial degradation time point is extracted as the degradation stage data.

[0056] The health index is used to represent the sensor signal data to describe the degradation process of the equipment, and then the OSW-PI-ACED method is used to determine the initial degradation time according to the constructed health index. Compared with the method of directly using the OSW-PI-ACED method to determine the initial degradation time according to the sensor signal data, the health index constructed in the embodiment of the present application can filter out invalid data, concisely and clearly express the initial degradation time and degradation process of the equipment, and can also reduce the interference caused by random noise in the process of determining the initial degradation time, thereby improving the accuracy of the health status prediction model and improving the efficiency of equipment health management.

[0057] 205. The rapid degradation stage data and the slow degradation stage data are respectively used to construct models to obtain a health status prediction model corresponding to the rapid degradation type and a health status prediction model corresponding to the slow degradation type.

[0058] In an embodiment of the present application, the equipment health management system obtains an MBR migration model corresponding to a rapid degradation type, an LP migration model corresponding to a rapid degradation type, an MBR migration model corresponding to a slow degradation type, and an LP migration model corresponding to a slow degradation type. It should be noted that while obtaining the migration model, the parameters of the MBR and LP migration models are set, and the parameters include learning rate, batch, initialization bias, weight, etc. Next, the equipment health management system inputs the rapid degradation stage data into the MBR model corresponding to the rapid degradation type to obtain the rapid degradation feature, and inputs the slow degradation stage data into the MBR model corresponding to the slow degradation type to obtain the slow degradation feature. Subsequently, the equipment health management system inputs the rapid degradation feature into the LP model corresponding to the rapid degradation type to obtain the health status label of the rapid degradation type, and inputs the slow degradation feature into the LP model corresponding to the slow degradation type to obtain the health status label of the slow degradation type.

[0059] Then, the equipment health management system uses the rapidly degraded health status label to calculate the first loss function, and uses the slowly degraded health status label to calculate the second loss function. Specifically, the equipment health management system uses the loss function calculation formula to calculate the rapidly degraded health status label to obtain the first loss function, and the loss function calculation formula is as follows:

[0060] Formula 2: L total =λL RUL +μL MMD

[0061]

[0062] Among them, y i is the real health status label, is the health status label of the rapid degradation type, λ is the first weight parameter, μ is the second weight parameter, n is the number of samplings in the degradation stage, D S is the source domain feature, D T is the target domain feature, n S The source domain d S The sample size, n T The target domain d T The sample size of ‖·‖ H is the reproducing kernel Hilbert space, is the calculation function for mapping data to the reproducing kernel Hilbert space. It should be noted that the calculation process of the second loss function is the same as the calculation process of the first loss function. RUL It is the mean of the squared difference between the actual health status label and the predicted health status label obtained by the model, which is part of the loss function.

[0063] Then, based on the back propagation algorithm, the equipment health management system uses the first loss function to update the network parameters of the MBR migration model corresponding to the rapid degradation type and the LP migration model corresponding to the rapid degradation type, and uses the second loss function to update the network parameters of the MBR migration model corresponding to the slow degradation type and the LP migration model corresponding to the slow degradation type. The training of the health status prediction model is completed by continuously reducing the gap between the actual health status label and the predicted health status label, that is, continuously reducing the loss function value.

[0064] Finally, the equipment health management system updates the network parameters of the MBR migration model corresponding to the rapid degradation type, the LP migration model corresponding to the rapid degradation type, the MBR migration model corresponding to the slow degradation type, and the LP migration model corresponding to the slow degradation type according to the preset number of iterations to obtain the health status prediction model corresponding to the rapid degradation type and the health status prediction model corresponding to the slow degradation type.

[0065] 206. The test data set is processed using a bidirectional long short-term memory convolutional neural network model and a Pearson correlation coefficient to obtain a second health indicator. The specified degradation stage data is extracted from the test data set based on the degradation stage detection method and the second health indicator, and the health status prediction model is tested using the specified degradation stage data.

[0066] In the embodiment of the present application, the equipment health management system uses a bidirectional long short-term memory convolutional neural network model and a Pearson correlation coefficient to process the test data set to obtain a second health indicator. It should be noted that the process of constructing the second health indicator is the same as the process of constructing the first health indicator.

[0067] If the equipment is determined to be degraded according to the second health indicator detection, the degradation type of the equipment is determined, and the specified degradation stage data is extracted from the test data set based on the degradation stage detection method. It should be noted that the process of extracting the degradation stage data based on the degradation stage detection method is the same as the process of extracting the rapid degradation stage data from the rapid degradation data set and the slow degradation stage data from the slow degradation data set using the degradation stage detection method. The health state prediction model corresponding to the degradation type is obtained, and the specified degradation stage data is input into the health state prediction model to obtain the health state prediction result output by the health state prediction model. Then, the equipment health management system obtains the true health state result of the equipment in the sensor signal sample data, and uses the health state prediction result and the true health state result to calculate the model accuracy; if the model accuracy is greater than or equal to the preset accuracy threshold, it means that the accuracy of the trained health state prediction model meets the requirements, and the health state prediction model is stored, which can ensure the accuracy and accuracy of the model, thereby realizing efficient management of equipment health.

[0068] Based on the above process, two types of degradation health status prediction models can be trained. Various data of the entire process from health to failure of the equipment are collected through sensors, and after degradation process classification and degradation stage data extraction, they are input into the MBR-LP migration model to obtain the equipment health status prediction results. In the case of small samples, the cross-operating health status prediction accuracy of 12 groups of data can reach more than 75%, which is much higher than the national standard requirement of 60%, indicating the advancement and superiority of the equipment health management method based on small sample data and adaptive migration proposed in this application.

[0069] 207. Obtain sensor signal data of the target device, and construct a health indicator of the target device based on the sensor signal data.

[0070] In an embodiment of the present application, the equipment health management system obtains a preset sampling interval, and collects data from the target device multiple times through the sensor according to the preset sampling interval to obtain sensor signal data. For example, the sampling time is set to 2S, the sampling interval is 1min, and the entire degradation process of the target device requires N samplings. Assuming that the data collected within 2S is a group of data, then N groups of data are obtained after N samplings, and these N groups of data are sensor signal data. Next, the equipment health management system obtains a pre-built bidirectional long short-term memory convolutional neural network model, inputs the sensor signal data into the bidirectional long short-term memory convolutional neural network model, and obtains the feature data set output by the bidirectional long short-term memory convolutional neural network model. For example, inputting the above-obtained N groups of data into the BLCAE model will obtain N feature data, and these N feature data are feature data set A, where A = (n1, n2, ...nm), and m is the number of samplings of the target device from the beginning to failure. Subsequently, the equipment health management system obtains a preset range, selects multiple feature data from the feature data set according to the preset range, and uses the average value of the multiple feature data as the benchmark data. Since the sensor signal data is collected in chronological order, after the features are extracted by the BLCAE model, the features in the feature data set are also in chronological order. Therefore, multiple feature data are directly selected in the feature data set according to the preset range. Among them, the preset range is the first 10%-20% of the data feature set. The reason is that the new parts are not run-in when they are first used, and the initial vibration may be large, so selecting features from the first data will result in the constructed health index being inaccurate. Finally, the equipment health management system obtains the Pearson correlation coefficient, and uses the Pearson correlation coefficient to calculate the correlation between each feature data and the benchmark data in chronological order to obtain the health index. It should be noted that the construction process of the health index is the same as the construction process of the first health index and the second health index mentioned above, and will not be described in detail here. The features are extracted by the bidirectional long short-term memory convolutional neural network model, and then the Pearson correlation coefficient is used to obtain the correlation between the features to obtain the health index, which can adapt to the sensor data of different components and different working conditions. It can not only accurately judge the health status of the equipment, but also improve the stability, extensiveness and accuracy of the model training, thereby realizing the equipment health management under small samples.

[0071] 208. Detect whether the target device is degraded according to the health indicator. If the target device is determined to be degraded according to the health indicator detection, determine the degradation type of the target device, and extract degradation stage data from the sensor signal data based on the degradation stage detection method and the health indicator.

[0072] In an embodiment of the present application, the equipment health management system detects whether the target equipment has degraded based on the health indicator. If the target equipment is determined to have degraded based on the health indicator detection, the degradation type of the target equipment is determined. Next, the equipment health management system extracts degradation stage data from the sensor signal data using the degradation stage detection method and the health indicator. It should be noted that the process of extracting degradation stage data from the sensor signal data based on the degradation stage detection method is the same as the process of extracting rapid degradation stage data from the rapid degradation data set and extracting slow degradation stage data from the slow degradation data set using the degradation stage detection method described above.

[0073] Specifically, the equipment health management system uses the 2sigma principle to extract the target health indicator from the health indicator, and uses the target health indicator to determine the detection threshold. Then, the equipment health management system uses an overlapping sliding window to slide over the target health indicator, and extracts the mutation point of the target health indicator from the multiple data points included in the overlapping sliding window based on the mutation point detection algorithm during the sliding process, and obtains multiple mutation points corresponding to each overlapping sliding window. For each overlapping sliding window, the equipment health management system uses the multiple data points included in the overlapping sliding window to determine multiple inflection points. Specifically, for each data point, the target data point is determined from the multiple data points according to the data point, wherein the target data point is the third data point behind the data point. Then, the equipment health management system calculates the slope value of the data point and the target data point; if the slope value is less than the preset threshold, it means that the data point may be a degradation point, and the data point is used as an inflection point, wherein the preset threshold is -0.002. Then, the equipment health management system sorts the multiple mutation points and the multiple inflection points in chronological order to obtain a time series, and extracts multiple target time points in the time series according to the principle of continuous excess detection verification. Specifically, the equipment health management system starts detection from the first time point in the time series according to the principle of continuous exceedance detection verification; if the i-th time point in the time series is greater than the detection threshold, it continues to detect the i+1-th time point until the i+j-th time point is detected, where the calculation formula for the parameter j is the above formula 1.

[0074] Subsequently, the equipment health management system sorts multiple target time points in chronological order to obtain the sorting results, and reads the first target time point in the sorting results as the initial degradation time. Finally, the equipment health management system extracts the degradation stage data from the sensor signal data according to the initial degradation time. Based on the degradation stage detection method, the equipment health management system can quickly and accurately extract the degradation stage data of the target equipment, thereby realizing the health status detection of different components and different working conditions of the same component, and realizing the health management of the equipment.

[0075] 209. Obtain a health status prediction model corresponding to the degradation type, input the degradation stage data into the health status prediction model, and obtain a health status prediction result output by the health status prediction model.

[0076] In an embodiment of the present application, the equipment health management system obtains a health status prediction model corresponding to the degradation type, inputs the degradation stage data into the health status prediction model, and obtains the health status prediction result output by the health status prediction model. In summary, the equipment health management system constructs reasonable health indicators for sensor data through BLCAE and Pearson correlation coefficient. Secondly, the initial degradation start time of the equipment is determined by the degradation stage detection method of OSW-PI-ACED, and the degradation stage data of each data is extracted. Finally, migration is performed between the same degradation categories so that degradation information can be effectively shared between similar fields or tasks, while avoiding negative migration and improving the reusability of historical knowledge and information. Compared with other methods, it shows that the present application can achieve higher accuracy with small samples, avoids the high cost of data preparation in certain specific devices, and performs better and more stable in cross-domain performance.

[0077] Based on the above process, a health status prediction schematic diagram proposed in the embodiment of the present application is as follows:

[0078] like Figure 2BAs shown in the figure, the original sensor signal data, i.e., the vibration data in the bearing degradation process, is obtained, and the original sensor signal data is divided into labeled source domain data and unlabeled target domain data. All source domain data are used as training sets, and the target domain data are randomly divided into two parts, of which 80% are training sets and 20% are test sets. The training set is used to construct health indicators, and then classification is performed. The health indicators can describe the degradation process of the equipment. Some equipment takes a long time from the beginning of degradation to failure, which can be called slow degradation; while some only take a short time, which can be called rapid degradation. The classification process is to divide the original data into rapid degradation or slow degradation categories according to the degradation process described by the health indicator. Then, a degradation stage detection method based on overlapping sliding windows-Pettitt mutation points, inflection point extraction-adaptive continuous exceedance detection verification is used to determine the initial degradation start time of the equipment, and the degradation stage data of each data in the data set is extracted. Subsequently, in order to improve the performance of the model in predicting the health status and health management of the equipment under small samples, the model is trained in the following way: using the MBR model to extract features; inputting the features into the LP model to obtain the health status label; using the health status label to calculate the loss function, and updating the network parameters of each layer of the MBR and LP models through the back propagation algorithm until the set number of iterations is reached, then the training is stopped to obtain the health status prediction model. This application proposes a device health management method for small sample data classification degradation and domain adaptive migration. By constructing health indicators, it reduces the interference of random noise in the search process of the equipment degradation stage, and sets adaptive thresholds and adaptive detection times for different data to make the model more generalized. After completing the data classification, migration is performed between the same degradation categories to reduce data distribution differences and avoid the occurrence of negative migration. At the same time, the health status prediction and health management of the equipment under small samples are completed.

[0079] The method provided in the embodiment of the present application obtains the sensor signal data of the target device, constructs the health index of the target device based on the sensor signal data, and detects whether the target device has degraded according to the health index; if the target device is determined to be degraded according to the health index detection, the degradation type of the target device is determined, and the degradation stage data is extracted from the sensor signal data based on the degradation stage detection method and the health index; the health state prediction model corresponding to the degradation type is obtained, and the degradation stage data is input into the health state prediction model to obtain the health state prediction result output by the health state prediction model. First, a reasonable health index is constructed for the sensor data through BLCAE (bidirectional long short-term memory convolutional neural network model) and Pearson correlation coefficient (Pearson correlation coefficient); secondly, the initial degradation start time of the device is determined by the degradation stage detection method of OSW-PI-ACED (based on overlapping sliding window-Pettitt mutation point, inflection point extraction-adaptive continuous exceedance detection verification), and the degradation stage data of each data is extracted; finally, migration is performed between the same degradation categories, so that the degradation information can be effectively shared between similar fields or tasks, while avoiding negative migration and improving the reusability of historical knowledge and information. By performing necessary processing on the limited data collected by the sensor, the data is classified to reduce the feature distribution differences between the source domain and the target domain, and domain adaptive migration is performed between the same categories, so that degradation information can be shared between similar fields or tasks to avoid the occurrence of negative migration, and the degradation stage data of the equipment can be accurately extracted to reduce the interference of irrelevant information on the prediction of the health status of the equipment. Even in the case of small samples, degradation information can be captured from a small amount of data, reducing data cross-domain deviation, improving the accuracy and stability of the model, and realizing health management of the equipment.

[0080] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a device health management device based on small sample data and adaptive migration, such as Figure 3A As shown, the device includes: a construction module 301, a detection module 302, a determination module 303 and a prediction module 304.

[0081] A construction module 301 is used to obtain sensor signal data of a target device and construct a health indicator of the target device based on the sensor signal data;

[0082] A detection module 302, configured to detect whether the target device is degraded according to the health indicator;

[0083] An extraction module 303, configured to determine the degradation type of the target device if it is determined that the target device is degraded according to the health indicator detection, and extract degradation stage data from the sensor signal data based on the degradation stage detection method and the health indicator;

[0084] The prediction module 304 is used to obtain a health status prediction model corresponding to the degradation type, input the degradation stage data into the health status prediction model, and obtain a health status prediction result output by the health status prediction model.

[0085] In a specific application scenario, the construction module 301 is used to obtain a preset sampling interval time, and collect data from the target device multiple times through the sensor according to the preset sampling interval to obtain the sensor signal data; obtain a pre-constructed bidirectional long short-term memory convolutional neural network model, input the sensor signal data into the bidirectional long short-term memory convolutional neural network model, and obtain a feature data set output by the bidirectional long short-term memory convolutional neural network model; obtain a preset range, select multiple feature data from the feature data set according to the preset range, and use the average value of the multiple feature data as the baseline data; obtain the Pearson correlation coefficient, and use the Pearson correlation coefficient to calculate the correlation between each feature data and the baseline data in chronological order to obtain the health indicator.

[0086] In a specific application scenario, the extraction module 303 is used to obtain the 2sigma principle, use the 2sigma principle to extract the target health indicator from the health indicator, and use the target health indicator to determine the detection threshold; obtain the overlapping sliding window and mutation point detection algorithm of the degradation stage detection method, use the overlapping sliding window to slide over the target health indicator, and extract the mutation point of the target health indicator from the multiple data points included in the overlapping sliding window based on the mutation point detection algorithm during the sliding process to obtain multiple mutation points corresponding to each overlapping sliding window; for each overlapping sliding window, use the multiple data points included in the overlapping sliding window to determine multiple inflection points, sort the multiple mutation points and the multiple inflection points in chronological order to obtain a time series; obtain the continuous exceedance detection verification principle, extract multiple target time points from the time series according to the continuous exceedance detection verification principle, sort the multiple target time points in chronological order to obtain a sorting result; read the target time point ranked first in the sorting result as the initial degradation time of the target device, and extract the degradation stage data from the sensor signal data according to the initial degradation time.

[0087] In a specific application scenario, the extraction module 303 is used to determine, for each of the data points, a target data point among the multiple data points based on the data point, where the target data point is the third data point behind the data point; obtain a preset threshold value, and calculate the slope value between the data point and the target data point; if the slope value is less than the preset threshold value, the data point is taken as an inflection point; and perform calculation processing on each of the data points in the overlapping sliding window to obtain the multiple inflection points.

[0088] In a specific application scenario, the extraction module 303 is used to start detection from the first time point of the time series according to the continuous exceedance detection verification principle; if the i-th time point in the time series is greater than the detection threshold, continue to detect the i+1-th time point until the i+j-th time point is detected, wherein,

[0089]

[0090] Wherein, n is the number of time points in the time series; when it is detected that there are j-1 time points greater than the detection threshold, the i-th time point is taken as the target time point; detection is continued according to the continuous exceedance detection verification principle until the last time point of the time series, to obtain the multiple target time points.

[0091] In specific application scenarios, such as Figure 3B As shown, the device also includes: a training module 305.

[0092] The training module 305 is used to obtain sensor signal sample data, obtain sample data with labels from the sensor signal sample data as source domain data, and obtain sample data without labels from the sensor signal sample data as target domain data, wherein the sensor signal sample data is equipment timing degradation data under different working conditions; obtain a preset ratio, divide the target domain data into a training data set and a test data set according to the preset ratio, and use the source domain data and the training data set as target training data sets; obtain a pre-constructed bidirectional long short-term memory convolutional neural network model and a Pearson correlation coefficient, use the bidirectional long short-term memory convolutional neural network model and the Pearson correlation coefficient to process the target training data set to obtain a first health indicator; and classify the first health indicator according to the degradation type. The indicators are divided into health indicators corresponding to rapid degradation types and health indicators corresponding to slow degradation types, and the degradation types include the rapid degradation type and the slow degradation type. Based on the health indicators corresponding to the rapid degradation type and the health indicators corresponding to the slow degradation type, the target training data set is divided into a rapid degradation data set and a slow degradation data set; a degradation stage detection method is obtained, and the rapid degradation stage data is extracted from the rapid degradation data set by using the degradation stage detection method, and the slow degradation stage data is extracted from the slow degradation data set by using the degradation stage detection method; the rapid degradation stage data and the slow degradation stage data are respectively used for model construction to obtain a health status prediction model corresponding to the rapid degradation type and a health status prediction model corresponding to the slow degradation type.

[0093] In a specific application scenario, the training module 305 is used to obtain the MBR migration model corresponding to the rapid degradation type, the LP migration model corresponding to the rapid degradation type, the MBR migration model corresponding to the slow degradation type, and the LP migration model corresponding to the slow degradation type; input the rapid degradation stage data into the MBR model corresponding to the rapid degradation type to obtain the rapid degradation feature, input the slow degradation stage data into the MBR model corresponding to the slow degradation type to obtain the slow degradation feature; input the rapid degradation feature into the LP model corresponding to the rapid degradation type to obtain the health status label of the rapid degradation type, input the slow degradation feature into the LP model corresponding to the slow degradation type to obtain the health status label of the slow degradation type; use the health status label of the rapid degradation type to calculate the first loss function, use the health status label of the slow degradation type A second loss function is calculated based on the health status label; a back propagation algorithm is obtained, and based on the back propagation algorithm, the first loss function is used to update the network parameters of the MBR migration model corresponding to the rapid degradation type and the LP migration model corresponding to the rapid degradation type; based on the back propagation algorithm, the second loss function is used to update the network parameters of the MBR migration model corresponding to the slow degradation type and the LP migration model corresponding to the slow degradation type; a preset number of iterations is obtained, and the network parameters of the MBR migration model corresponding to the rapid degradation type, the LP migration model corresponding to the rapid degradation type, the MBR migration model corresponding to the slow degradation type, and the LP migration model corresponding to the slow degradation type are updated according to the preset number of iterations to obtain a health status prediction model corresponding to the rapid degradation type and a health status prediction model corresponding to the slow degradation type.

[0094] In a specific application scenario, the training module 305 is used to obtain a loss function calculation formula, and use the loss function calculation formula to calculate the health status label of the rapid degradation type to obtain the first loss function, wherein,

[0095] L total =λL RUL +μL MMD

[0096]

[0097] Among them, y i is the real health status label, is the health status label of the rapid degradation type, λ is the first weight parameter, μ is the second weight parameter, n is the number of sampling times in the degradation stage, D S is the source domain feature, D T is the target domain feature, n SThe source domain d S The sample size, n T The target domain d T The sample size of ‖·‖ H is the reproducing kernel Hilbert space, is a computational function that maps data to the reproducing kernel Hilbert space.

[0098] In a specific application scenario, the training module 305 is used to process the test data set using the bidirectional long short-term memory convolutional neural network model and the Pearson correlation coefficient to obtain a second health indicator; if the device is determined to be degraded based on the detection of the second health indicator, the degradation type of the device is determined, and a health status prediction model corresponding to the degradation type is obtained, and the specified degradation stage data is extracted from the test data set based on the degradation stage detection method and the second health indicator; the specified degradation stage data is input into the health status prediction model to obtain a health status prediction result output by the health status prediction model; the true health status result of the device is obtained from the sensor signal sample data, and the model accuracy is calculated using the health status prediction result and the true health status result; a preset accuracy threshold is obtained, and if the model accuracy is greater than or equal to the preset accuracy threshold, the health status prediction model is stored.

[0099] The device provided in the embodiment of the present application obtains the sensor signal data of the target device, constructs the health index of the target device based on the sensor signal data, and detects whether the target device has degraded according to the health index; if the target device is determined to be degraded according to the health index detection, the degradation type of the target device is determined, and the degradation stage data is extracted from the sensor signal data based on the degradation stage detection method and the health index; the health state prediction model corresponding to the degradation type is obtained, and the degradation stage data is input into the health state prediction model to obtain the health state prediction result output by the health state prediction model. First, a reasonable health index is constructed for the sensor data through BLCAE (bidirectional long short-term memory convolutional neural network model) and Pearson correlation coefficient (Pearson correlation coefficient); secondly, the initial degradation start time of the device is determined by the degradation stage detection method of OSW-PI-ACED (based on overlapping sliding windows-Pettitt mutation points, inflection point extraction-adaptive continuous exceedance detection verification), and the degradation stage data of each data is extracted; finally, migration is performed between the same degradation categories, so that the degradation information can be effectively shared between similar fields or tasks, while avoiding negative migration and improving the reusability of historical knowledge and information. By performing necessary processing on the limited data collected by the sensor, the data is classified to reduce the feature distribution differences between the source domain and the target domain, and domain adaptive migration is performed between the same categories, so that degradation information can be shared between similar fields or tasks to avoid the occurrence of negative migration, and the degradation stage data of the equipment can be accurately extracted to reduce the interference of irrelevant information on the prediction of the health status of the equipment. Even in the case of small samples, degradation information can be captured from a small amount of data, reducing data cross-domain deviation, improving the accuracy and stability of the model, and realizing health management of the equipment.

[0100] It should be noted that for other corresponding descriptions of the functional units involved in the device health management device based on small sample data and adaptive migration provided in the embodiment of the present application, please refer to Figure 1 and FIG. 2A to FIG. 2B The corresponding description in will not be repeated here.

[0101] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0102] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

[0104] In an exemplary embodiment, see Figure 4 , and also provides a computer device, which includes a bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device, wherein each functional unit can communicate with each other through the bus. The memory stores a computer program, and the processor is used to execute the program stored in the memory and execute the device health management method based on small sample data and adaptive migration in the above embodiment.

[0105] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the device health management method based on small sample data and adaptive migration.

[0106] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by hardware, or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0107] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present application.

[0108] Those skilled in the art will appreciate that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more devices different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0109] The above application serial numbers are for description only and do not represent the advantages or disadvantages of the implementation scenarios.

[0110] The above disclosure only discloses several specific implementation scenarios of the present application. However, the present application is not limited thereto, and any changes that can be conceived by technicians in this field should fall within the scope of protection of the present application.

Claims

1. A device health management method based on small sample data and adaptive migration, characterized in that: include: Acquire sensor signal data of a target device, and construct a health indicator of the target device based on the sensor signal data; Detecting whether the target device is degraded according to the health indicator; If the target device is determined to be degraded according to the health indicator detection, then determining the degradation type of the target device, and extracting degradation stage data from the sensor signal data based on the degradation stage detection method and the health indicator; A health status prediction model corresponding to the degradation type is obtained, and the degradation stage data is input into the health status prediction model to obtain a health status prediction result output by the health status prediction model.

2. The method according to claim 1, characterized in that The acquiring sensor signal data of the target device and constructing the health indicator of the target device based on the sensor signal data includes: Obtaining a preset sampling interval time, and collecting data of the target device multiple times through a sensor according to the preset sampling interval to obtain the sensor signal data; Obtain a pre-built bidirectional long short-term memory convolutional neural network model, input the sensor signal data into the bidirectional long short-term memory convolutional neural network model, and obtain a feature data set output by the bidirectional long short-term memory convolutional neural network model; Acquire a preset range, select multiple feature data from the feature data set according to the preset range, and use an average value of the multiple feature data as benchmark data; The Pearson correlation coefficient is obtained, and the correlation between each characteristic data and the benchmark data is calculated in sequence using the Pearson correlation coefficient in chronological order to obtain the health index.

3. The method according to claim 1, characterized in that: The step of extracting degradation stage data from the sensor signal data based on the degradation stage detection method and the health indicator includes: Acquire the 2sigma principle, extract the target health indicator from the health indicators by using the 2sigma principle, and determine the detection threshold by using the target health indicator; Obtaining an overlapping sliding window and a mutation point detection algorithm of the degradation stage detection method, sliding the overlapping sliding window over the target health indicator, extracting the mutation point of the target health indicator from a plurality of data points included in the overlapping sliding window based on the mutation point detection algorithm during the sliding process, and obtaining a plurality of mutation points corresponding to each overlapping sliding window; For each of the overlapping sliding windows, multiple data points included in the overlapping sliding window are used to determine multiple inflection points, and the multiple mutation points and the multiple inflection points are sorted in chronological order to obtain a time series; Acquire a continuous excess detection and verification principle, extract multiple target time points in the time series according to the continuous excess detection and verification principle, sort the multiple target time points in chronological order, and obtain a sorting result; The first target time point in the sorting result is read as the initial degradation time of the target device, and the degradation stage data is extracted from the sensor signal data according to the initial degradation time.

4. The method according to claim 3, characterized in that The method of using the multiple data points included in the overlapping sliding window to determine the multiple inflection points includes: For each of the data points, determining a target data point among the multiple data points according to the data point, the target data point being the third data point behind the data point; Obtaining a preset threshold value, and calculating a slope value between the data point and the target data point; If the slope value is less than the preset threshold, the data point is taken as an inflection point; Calculation is performed on each of the data points in the overlapping sliding windows to obtain the multiple inflection points.

5. The method according to claim 3, characterized in that: Extracting multiple target time points in the time series according to the continuous exceedance detection verification principle includes: Starting detection from the first time point of the time series according to the continuous exceedance detection verification principle; If the i-th time point in the time series is greater than the detection threshold, then continue to detect the i+1-th time point until the i+j-th time point is detected, where: Wherein, n is the number of time points in the time series; When it is detected that there are j-1 time points greater than the detection threshold, the i-th time point is used as the target time point; The detection is continued according to the continuous exceedance detection verification principle until the last time point of the time series, to obtain the multiple target time points.

6. The method according to claim 1, characterized in that Before obtaining the health status prediction model corresponding to the degradation type, the method further includes: Acquire sensor signal sample data, acquire sample data with labels from the sensor signal sample data as source domain data, acquire sample data without labels from the sensor signal sample data as target domain data, wherein the sensor signal sample data is equipment timing degradation data under different working conditions; Obtaining a preset ratio, dividing the target domain data into a training data set and a test data set according to the preset ratio, and using the source domain data and the training data set as a target training data set; Obtaining a pre-built bidirectional long short-term memory convolutional neural network model and a Pearson correlation coefficient, and using the bidirectional long short-term memory convolutional neural network model and the Pearson correlation coefficient to process the target training data set to obtain a first health indicator; dividing the first health indicator into a health indicator corresponding to a rapid degradation type and a health indicator corresponding to a slow degradation type according to the degradation type, wherein the degradation type includes the rapid degradation type and the slow degradation type; Dividing the target training data set into a rapid degradation data set and a slow degradation data set based on the health indicator corresponding to the rapid degradation type and the health indicator corresponding to the slow degradation type; Acquire a degradation stage detection method, extract rapid degradation stage data from the rapid degradation data set by using the degradation stage detection method, and extract slow degradation stage data from the slow degradation data set by using the degradation stage detection method; The rapid degradation stage data and the slow degradation stage data are respectively used to construct models to obtain a health status prediction model corresponding to the rapid degradation type and a health status prediction model corresponding to the slow degradation type.

7. The method according to claim 6, characterized in that The model construction is performed by respectively using the rapid degradation stage data and the slow degradation stage data to obtain a health status prediction model corresponding to the rapid degradation type and a health status prediction model corresponding to the slow degradation type, including: Acquire an MBR migration model corresponding to the rapid degradation type, an LP migration model corresponding to the rapid degradation type, an MBR migration model corresponding to the slow degradation type, and an LP migration model corresponding to the slow degradation type; Inputting the rapid degradation stage data into the MBR model corresponding to the rapid degradation type to obtain rapid degradation characteristics, and inputting the slow degradation stage data into the MBR model corresponding to the slow degradation type to obtain slow degradation characteristics; Inputting the rapid degradation feature into the LP model corresponding to the rapid degradation type to obtain the health status label of the rapid degradation type, and inputting the slow degradation feature into the LP model corresponding to the slow degradation type to obtain the health status label of the slow degradation type; The first loss function is calculated by using the health status label of the rapid degradation type, and the second loss function is calculated by using the health status label of the slow degradation type; Acquire a back propagation algorithm, and based on the back propagation algorithm, use the first loss function to update network parameters of an MBR migration model corresponding to the rapid degradation type and an LP migration model corresponding to the rapid degradation type; Based on the back propagation algorithm, the second loss function is used to update the network parameters of the MBR migration model corresponding to the slow degradation type and the LP migration model corresponding to the slow degradation type; Obtain a preset number of iterations, and update network parameters of the MBR migration model corresponding to the rapid degradation type, the LP migration model corresponding to the rapid degradation type, the MBR migration model corresponding to the slow degradation type, and the LP migration model corresponding to the slow degradation type according to the preset number of iterations to obtain a health status prediction model corresponding to the rapid degradation type and a health status prediction model corresponding to the slow degradation type.

8. The method according to claim 7, characterized in that The method of calculating a first loss function using the health status label of the rapid degradation type includes: Obtain a loss function calculation formula, and use the loss function calculation formula to calculate the health status label of the rapid degradation type to obtain the first loss function, wherein, L total =λL RUL +μL MMD Among them, y i is the real health status label, is the health status label of the rapid degradation type, λ is the first weight parameter, μ is the second weight parameter, n is the number of sampling times in the degradation stage, D s is the source domain feature, D T is the target domain feature, n S The source domain d S The sample size, n T The target domain d T The sample size of ||·|| H is the reproducing kernel Hilbert space, is a computational function that maps data to the reproducing kernel Hilbert space.

9. The method according to claim 6, characterized in that The method further comprises: Processing the test data set using the bidirectional long short-term memory convolutional neural network model and the Pearson correlation coefficient to obtain a second health indicator; If it is determined that the device is degraded according to the second health indicator detection, then determining the degradation type of the device, and extracting the designated degradation stage data from the test data set based on the degradation stage detection method and the second health indicator; Acquire a health status prediction model corresponding to the degradation type, input the specified degradation stage data into the health status prediction model, and obtain a health status prediction result output by the health status prediction model; Acquire the true result of the health status of the device from the sensor signal sample data, and calculate the model accuracy using the health status prediction result and the true result of the health status; A preset accuracy threshold is obtained, and if the model accuracy is greater than or equal to the preset accuracy threshold, the health status prediction model is stored.

10. A device health management device based on small sample data and adaptive migration, characterized in that: include: A construction module, used to obtain sensor signal data of a target device, and construct a health indicator of the target device based on the sensor signal data; A detection module, used to detect whether the target device is degraded according to the health indicator; an extraction module, configured to determine the degradation type of the target device if it is determined that the target device has degraded according to the health indicator detection, and extract degradation stage data from the sensor signal data based on a degradation stage detection method and the health indicator; The prediction module is used to obtain a health status prediction model corresponding to the degradation type, input the degradation stage data into the health status prediction model, and obtain a health status prediction result output by the health status prediction model.

Images